DISTRIBUTED BY: National Technical Information Service U ... · The description of the new psychological test ... tions for official action normally are conveyed to appropriate military

AD-751 761

DEVELOPMENT AND EVALUATION OF A NEW ACBAND APTITUDE AREA SYSTEM

Milton H. Maier, et al

Army Behavior and Systems Research LaboratoryArlington, Virginia

September 1972

DISTRIBUTED BY:

National Technical Information ServiceU. S. DEPARTMENT OF COMMERCE5285 Port Royal Road, Springfield Va. 22151

* i m l.m • m

Tchnical Research Note 239 A1

to i DEVELOPMENT AND EVALUATION OF A NEW ACBg AND APTITUDE AREA SYSTEM

Milton H. Maier and Edmund F. Fuchs

MILITARY SELECTION RESEARCH DIVISION

Reproduced byJIjNATIONAL TECHNICALINFORMATION SERVICE -LI/,,.;IL_.,....Ii.- J U L,.

U S Deportment of Commerce

Sprnngfield VA 22153

U. S. ArmyBehavior and Systems Research Laboratory

0 September 1972

Approved for Public release, distribution unlimited.

BEHAVIOR AND SYSTEMS RESEARCH LABORATORYAn activity of the Chief, Research and Development

J. E. UHLANERDirector 3

.• ..... ;. H h... .............

ff ... ................

DlSfl3TIDI/AV~uL•It91ITY CODES

NOTICES

DISTRIBUTION: Primary distribution of this report has been made by BESRL. Please addresscorrespondence concerning distribution of reports to: U. S. Army Behavior and Systems Research .

Laboratory, Attn: RDMR-BLZ, 1300 Wilson Boulevard, Arlington, Virginia 22209.

IFINAL DISPOSITION: This report may be destroyed when it is no longer needed. Please do not

return it to the Behavior and Systems Research Laboratory.

NOTE: The findings in this report are not to be construed as an official Department of the Army

position, unless so designated by other authorized documents.

I

UnclassifiedSecurity Claszification .__ ,_ ,_ _ _ _

DOCUMENT CONTROL DATA -R &D(SCurtfy ctassili, .. litn of title, body of abstract and Indexing annotation must be entered when ,t j overatl report is clas•lfed;

I. ORIGINATING AC IIYIIY (Csotorste aunthor) Ize. REPORT SCCURITY CLASSIFICATION

Behavior and Systems Research Laboratory, Unclassified

Arlington, 7irginia 2b. GROUP

3. REPORT TITLE

DEVELOPMENT AND EVALUATION OF A NEW ACB AND APTITUDE AREA SYSTEM

4. OESCRIPTIVE NOTES (Type of report and Inclusive dates)

S. AUTHOR(S) (First name, middle tritiat, last inme)

Milton H. Maier and Edmund F. Fuchs

6. REPORT DATE 7a. TOTAL NO. OF PAGES 7b. N(. OF REFS

September 1972 ____ f4 9". CONTRACT OR GRANT NO. 9a. ORIGINATOR'S I'ORT NUIBE R(:)

b. PiROJECT NO.

DA R&D Proj. No. 2Q062106A722 Technical Research Note 239c. 9b. OTHER REPORT NOIS) (Any othe• nubera that nay be assignad

Differential Classification this report)

d. b-l110. DISTRIBUTION STATEMENT

Approved for public release; distribution unlimil.ed

II. SUPPLEMENTARY NOTrs7 .12 SPONSORING MILITARY ACTIVITY

Deputy Chief of Staff forPersonnelWash.,DCUS Contin(!ntal Army Cmd., Ft Monroe, VA

13. AOSTRACT

Army personnel managers have a continuing need tc select, classify, and assign totraining and jobs large numbers of men who enter the service. Since the Army Classi-fication Batter, (ACB) is an integral part of the assignment process, accuracy ofscores has a significant influence on the appropriateness of assignment. BESRL'sDIFFERENTIAL CLASS:FICATION Work Unit has an ongoing research program to keep theclassification battery effective and up to date. As part of the overall effort,,a newACB and aptitude area system have been developed wlhich result in an improved system ofclassification for training and jobs. The description of the new psychological testbattery and aptitude areas and an assessment of their effectiveness in relation to theutilization and performancc of Army enlisted input is ',iven in Technical ResearchReport 1177. The present r,,blication deals with BESRL research Londucted to evaluatea large number of tests as predictors of success in the different groups of MilitaryOccupational Specialties (MOS) and to select tests for aptitude area composites.

Experimental tests and tests of the operational ACB (administered to about 25,000men in over 100 MOS training courses) were evaluated against performance in the train-ing courses. Validity coefficients of the variables with final course grades in theMOS courses were computed and corrected to reflect population values. Regressionequations for all tests were computed in each MOS sample, and for each MOS group asequence of test selections was performed to determine which tests contributed sig-

nificantly to validity. These statistical analyses resulted in a test battery of 16

measures and the formulation of 9 aptitude areas designated as selectors for 9 MOS

D fMSecurity Clasification

DD Form 1473

15. ABSTRACT - Continuedgroups. Each aptitude area consisted of from three to five tesLE, each

unit-weighted. Finally, through simulation runs, estimates were derivedof operational effects of introducing the new classification system.

Findings indicate the new Army Classification Battery and aptitudeareas to be superior tj the previous system. Average validity of the newaptitude area composites across all MOS groups is higher than that of cheprevious composites. Supporting statistical analyses are provided indetailed tables in appendixes to the Research Note.

I

Uncl issifiedSecurity Classification - - -A

4.LINK A LINK Ui L.NK C•KE[Y W10ROS ,. ,-

ROLE WT ROLE WT ROLE WT

SDifferential classification

Army Classification Battery

SAptitude areas

Classification sys:em

Aptitude area composite

-Aptitude area system* Military Occupational Specialty (MOS)|Military psychology

Psychological Measurement

Statistical analyses

Test validit"

Validity patterns

4

1

j?

ii 4

Secitylt Cl•ailticatlon

Technical Research Note 239 1i

2

DEVELOPMENT AND EVALUATION OF A NEW ACB

AND APTITUDE AREA SYSTEM

Milton H. Maier andEdmund F. Fuchs

MI L ITARY SELECTION RESEARCH DIVISIONEdmund F. Fuchs, Chief

BEHAVIOR AND SYSTEMS RESEARCH LABORATORY

Office, Chief of Research and DevelopmentDepartment of the Army

1300 Wilson Boulevard, Arlington, Virginia 22209

Septemter 1972

Army Projoct Number Difterential Classification b-l 120062106A122 •-.

Approved for public release; distribution unlimited

BESRL Technical research Reports and Technical Research Notes ale intended forsponsors of R&D tasks and other research and military agencies. Any findingsready for implementation at the time of publication are presented in the latter partof the Brief. Upon completion of a major phase of the task, formal recommenda-tions for official action normally are conveyed to appropriate military agenciesby briefing or Disposition Form.

I t

*21

%

FOREWORD

The DIFFERENTIAL CLASSIFICATION Work Unit applies psychological measurement meth-ods to enable the Army to make best use of the skills ant aptitudes of its enlisted personnelthrough increasingly accurate and differentiated measures of individual potential. Research isconducted to maintain and improve the effectiveness of the As-my Classification Battery and r,.,lated techniques and of conditions which may interact wi',i the classification z.sts and thJs

.affect the basis for utilization of the enlisted input--changes i i training programs and jctb co;tmntand environment, for example.

As part of the overall effort, a new Army Classification Battery and new aptitude areas have

been developed which result in an improved system of classification for training and j')bs. Thepresent Technical Research Note deals with BESRL research conducted to evaluate a large num-ber of tests as predictors of success in the different groups of Military Occupational Specialtiesand to select tests for aptitude area composites.

The entire research work unit is responsive to special requirements of the Deputy Chief ofStaff for Personnel and the U. S. Continental Army Command, as well as to objectives of ArmyRDT&E Project 20062106A722, "Selection and Behavioral Evaluation, "FY 1972 Work Program.

J. E. UHLANER, Director

Behavior and Systems

Research Laboratory VI

I!I/

-I

DEVELOPMENT AND EVALUATION OF A NEW ACB AND APTITUDE AREASYSTEM

BRIEF

Requirement:

To develop and evaluate a new Army Classification Battery (ACB) and new aptitude area com-posites for use in the selection of enlisted men and their classification to training and jobs.

Procedure:

- Experimental and operational tests made up a battery administered to about 25,000 men ir,over 100 Army Military Occupational Specialty (MOS) training courses. Validity coefficients ofexperimental and operational test variables with final course grades in the MOS courses were com-

puted. Coefficients were corrected to reflect population values. Regression equations for alltests were computed in each MOS sample, and MOS judged similar in job demands and showing

similar test validity patterns were combined. For each MOS group, a seauence of test selectionswas performed to determine which tests contributed significantly to validity. These statistical

analy:et resulted in a test battery of 16 measures and the formulation of nine aptitude areas des-ignated as selectors for nine MOS groups. Each aptitude aea consists of from three to five tests,each test being unit weighted. Finally, through simulation runs, ,'-1mates were made of opera-

tional effects of introducing the new classification system.

Findings:

Mean validity coefficit nt for the new ACB is .65, compared to .55 for the prior battery. A 20percent reduction in attrition in advanced individual training estimated as a result of the highervalidity.

IIn the simulation runs, overall level of predicted training performance was considerably higherwith the new system. Taking 100 as base score representing expected performance w:ider random

assignment, mean predicted criterion score with the new battery was 104.6 comparea to 102.6with the old battery.

A more equitable distribution of ability across MOS groups is attained witn' the new battery.In all job areas, predicted performance is above average, means running from 103 to 108.

With the new system, the number of men performing at superior levels is increased by 15percent;, the number of men performing at marginal levels is decreased by 20 percent.

.

Because general mental ab~lity is more strongly reflected ir. every aptitude area composite tharin the previous composites, fewer men of marginal mental level (10-30 AFQT score) will achiL-ve

one or more aptitude area scores of 90 or above. So long as aptitude area scores enter ir'o thescreening process, more men in AFQT Category IV (1O-jos) will need to be examined oefore a

specified number of men in Category IV are qualified for service. The men who do oualify underthe new system can be expected to do better in training and adapt better to Army life.

Utilization of Findings:

The new Aimy Classification Battery and aptitude area system is rd-oposed for implementation

in Calendar Year 1972.

S.4

.2

I÷-I

5"+

DEVELOPMENT AND EVALUATION OF A NEW ACB AND APTITUDE AREASYSTEM

CONTENTS

Page

BACKGROUND

DEVELOPMENT OF THE NEW APTITUDE AREA SYSTEM 2

The Experimental Tests 2

Sampling and Data Collection 2Data Analysis 5Grouping the MOS 6Test Selection 7Validity of Composites 9Validity of Education, Age, and Selected High SchoolCourses 14

EVALUATION OF THE NEW APTITUDE AREA SYSTEM 17

Reduction of Attrition 17Simulated Evaluation of Perfo-"mance 17Selection of Men with Marginal Mental Ability 25

CONCLUSION 28

LITERATURE C' TE D 29

APPENDIXES 51

DISTRIBUYION 50

DD Form 1473 (Document Control - R&D) 52

TABLES

Table 1. Description of experimental tests

2. Composition of MOS groups 10

I

% * -.

-

Tables (Cont) Pg

Page-

3. Content of new and prior Army classification batteries 12

4. New aptitude area composites 13

5. Validity of new ACB composites 15

6. Mean beta weights of level of education, age, andarithmetic reasoning 16

7. Mean predicted criterion scores under the newaptitude area system and under the previous system 19

8. Expected performance in old and new aptitude area

systems 21

9. Mean preIcced criterion scores with assignmentsmade by different weighting schemes 23

10. Distributior. of number of AQB aptitude area scoresat or above 90 for men in uental category IV 26

11. Distribution of number of new aptitude area scoresat or above 90 for full range of mental ability 27

DEVELOPMENT AND EVALUATION OF A NEW ACB AND A)TITUDEAREA SYSTEM

A new Army Classification Battery and a new aptitude area system havebeen developed to meet more effectively the needs of the modern Aray.Since the 1958 aptitude area system was introduced operetionally, tech-nological changes have greatly increased the complexity of Army jobs.Greater competence--in some cases a different kind of competence--is re-quired of the men who make up the Army's enlisted forces. Training pro-grams have been redesigned to develop the competence and skills required.Some Army schools training enlisted men for assignment in Military Occupa-tional Specialties (MOS) have asked for more comprehensive measures oftrainability sc that the men selected on the basis of a given aptitudearea prerequisite will have greater likelihood of successfully completingthe training.

To meet changing requirements, the Behavior and Systems Research Lab-oratory has conducted a large-scale research program to develop moreappropriate classification measures. Experimental and operational testswere administered to about 25,000 men in over 100 different trainingcourses, covering the gamut of openings available to newly enlisted men.The men came from all over the country and were representative of Armyinp'-it. The sampling of men and training courses provided a solid sci-entiiii base for developing a new Army Classification Battery and aptitudesystem.

Experimental tests and tests of the operational Army ClassificationBattery (ACB) were evaluated against performance in the training courses.The first test of a recruit's ability to meet Army demands occurs in hisjob training course, and job training grades are the fir8t objective as-sessment of how well individuals perform in their initial assignments.These initial assignments are made largely on the basis of ACB measures.For these purposes the accuracy with which training grades are predictedserves as a standard by which to evaluate the effectiveness of the clas-sification tests.

Operational test scores and background data were obtained from offi-cial records. The experimental tests oere administered to trainees in thecourses as the men began their instruction. The men in each course werefollowed through training to obtain measures of how well they succeededc in subsequent duty assignments. Analysis of these date will be presentedin a separate report.

l'1

DEVELOPMENT OF THE NEW APTITUDE' AREASYSTEM

The Experimental Tests

The experimental measures were designed to expand the test coverageof the operational ACB and to update certain of the component zests. In-cluded were measures of interest in a number of occupational areas. Newtests were tried out to e:xtend measurement in the important domain ofgeneral mental ability. Updated versions and modifications of existingtests were included in the domains of mechanical ability and perceptualability. The experimental tests are described in Table 1.

Each test in the experimental battery was designed to measure a setof skills, knowledge, or interests related to performance inma group ofMOS. Previous developmental research had shown that the experimental

test were valid for some MOS, but their differentiai!validity for variousoccupational areas and the extent of their unique validity remained tobe determined. The present analysis was designeed to 6etermine whethereach of the tests was an accurate predictor of success in a relevantgroup of MOS--for example, a test. of mechanical ability must be validfor mechanical maintenance MOS--and whether it makes a unique contribution

to prediction--that is, has a large beta weight. It should also haveits highest validity for related MOS, and lower validity for unrelapedMOS. Mechanical tests, for example, should be le~s valid for clerical-administrative jobs than for mechanical jobs. Tests of general abilitymay have validity for a wide variety of jobs because successful perfor-mance in most jobs requires a significant component of general mentalability. The requirement for differential validity therefore is not sostringent for these tests.

Sampling and Data Collection

The samples on which experimental and operational test scores andbackground data were collected have been described in'BESRL TechnicalResearch Report 1159!-.. The MOS training courses contained some_ studentswho had had extensive Army training and job expeeience. The training

performance of these men is probably of a different order than that ofmen who recently entered the Army. The samples were purified by de-leting all men who were known to have had other Army job assignments.Female trainees were also dropped from the samples. The populationto which the results of this research generalize is zom-osed of maleswho recently entered the Army and are in their initial training program.

Maier, M.H, and E.F., Fichs. Development of improved aptitude areacomposites for enlistrd classification. Technical Research Report

, (AD 351 268/ Behavior and Systems Research Laboratory,Arlington, VA. September lOW4,.

-2-

Table I

DESCRIPTION OF EXPERIMENTAL TESTS

1. Subtraction and Division (100 items). The task involves simplearithmetic computations; no reading is required.

2. Tool Knowledge (20 items'. The task is to identify the functionsof hand tools, or to identify which tool among the alternatives is

used in the same trade as the tool shown in the item. The item andalternatives are always shown in pictorial form, with the exception ofthe last item, which contains words in the alternatives.

3. Differential Classification Inventory. The task in one group ofitems is to indicate the activity the person thinks he is best at orwould rather do. In other items, the respondent is to rate the qual-ity of his performance or level cl interest, or to describe his per-ception of himself and of others. Four scores are obtained:

Electronics Interest Scale (20 items)

Mechanical Interest Scale (20 items)

Clerical Interest Scale (20 items)General Adjustment Scale (20 items)

4. Electronics Information '20 items). The task is to identify elec-trical and electronic equipment and components and to define electri-cal terms.

5. Mechanical Comprehension (23 items). Measures knowledge of me-chanical forces, valves, gears, and pulleys.

•. Mathematics Knowledge (20 items). Measures ability in algebra,geometry, and intcrpretation of graphs.

7. Science (20 items). Measures ability to read electrical diagramsidentify electrical components, and explain electrical circuits.

8. Personal Inventory for Electronics. Includes items on attitudesto,,ard authority figures, home and school background, frequency ofusing tools, activities engaged in,and personality items. Four itemsare in common, Two scores are derived:

High Skill (25 items)Low Skill (25 items)

'. Pattern Analysis '20 items). The task is to identify which offour forms results from folding a pattern.

Table ! continued

10. Science Knowledge (50 items). Measures knowledge of biclogy,chemistry, physiology, and medicine.

.1, Electronics Pictures (20 items). The task is to identify electri-cal components and various symbols used in electrical diagrams.

12. Occupational Interest Inventory. Examinee indicates degree ofliking for specific activities of the job and desired level of super-vision and skill requirements in a job. Scores are obtained on threeareas of interest:

Biological (20 items)Construction ('20 items)General Job '20 items)

The examinee also indicates whether he has had each of the follow-ing ten high school courses:

S~Bi olIogy Physic s

Chemistry Earth SciencesGeneral Science AlgebraElectrical Shop WoodworkHygiene Metal Work

15. Trade Information (25 items). Measures knowledge of electricalwiring, carpentry, construction equipment, hardware (as found in hard-ware store), and principles of design.

14. Attention-to-Detail (60 items). The task is to count the numberof C's, which varies from zero through four, in a series of O's andCIs.

Only the Subtraction and Division Test and the Attentlon-to-De-

tail Test are highly speeded. All the tests are scored rights only(R) except the Subtraction and Division Test, which is scored rightsminus one-fourth wrongs (R - W/4). All the tests except two involveextensive verbal comprehension. The Subtraction and Division Testcontains no words. The Tool Knowledge Test consists of pictures inthe items and alternatives with a few verbal statements to identifythe task for a group of items.

-4-

L

A large proportion of the trainees, about 25%, did not graduate onschedule, and course grades for this group, if available, were of ques-tionable meaning. The exceptional cases consisted of three main subgroups,academic failures, academic turnbacks or recycles, and withdrawals fornonacademic reasons. The withdrawals were assumed to be randomly dis-tributed and were dropped from further consideration. The failures andturnbacks were rot randomly distributed; had they been d&opped, the sam-ples would have been biased, as men with low training performance wouldhave been deleted. An analysis was conducted to determine how to assigngrades to turnbacks and failures-4 The results indicated that the scaleposition of failing grades was one standard deviation below minimum pass-ing, and all failures were assigned this grade. The scale position ofgrades for turnbacks was one-half a standard deviation below the meanof the graduates; turnbacks without grades were assigned that grade, andif a graae was reported for a turnback, that grade was the one used inthe analysis. With the inclusion of the academic failures and turnbacks,the sample for each training course had no known source of criterion bias.

Data Analysis

Validity coefficients for all tests were computed against final coursegrade in each MOS training course. Product-moment intercorrelation ma-trices of test scores and final course grade were computed for all MOSsamples. Because each sample had previously been selected on the basisof operational ACB scores, the coefficients were corrected by the multi-

variate restriction in iange formula to reflect the results that wouldhave been obtained for a random sample of the population of men eligiblefor Army service. The complete matrix of intercorrelations, corrected topopulation values, of all test, backgrou.nd, and training performance vari-

ables is shown in Table A-i of the Appencx.

At this stage, with 100 samples and more than 30 variables, over 3,000validity coefficients had been computed. Regression equations were com-puted in each sample to determine what aptitudes, knowledge, and interestswere important in identifying potential for success in each MOS. The re-suits were examined to determine MOS that had similar requirements. MOSjudged similar in terms of job demands and profiles of test validity weregrouped in order to obtain more reliable estimates of validity. Meanvalidity vectors were computed for each group of MOS with the MOS samplesunit-weighted rather than weighted by number of cases in each sample.Test selections were performed on the mean validity vectors to determinewhich tests had unique validity for each group of MOS. Finally, simula-

tion runs were made to obtain estimates of the impact on training perfor-

mance to be expected from the new classification system and to obtain es-timates of the distributions of aptitude area scores at d!fferentAFQT levels.

Maier, M.H. Procedures for assigning grades to failures and turnbacksin Army school courses, Technical Research Note Iq7. (AD f5 j5ý2).Behavior and Systems Research Laboratory, Arlington, VA. April 11qýC.

-5-

Grouping the MOS

A major problem was to find a reasonable basis for grouping the MOSinto homogeneous clusters. About 90 different MOS were included in theover 100 samples obtained at the training schools. Several hundred MOSare potentially available to a recruit entering the Army. Because ofquota restrictions, not all MOS are open at any one time. Even so, thenumber of possibilities is too great for the human mind to consider si-multaneously. Besides, most of the MOS fall into homogeneous clustersinvolving similar job tasks. Rational procedures were required to findways of reducing the complexity. Two considerations for combining MOSinto meaningful groups were empirical data an! operational convenience.Any grouping should combine cnly those MOS that are similar in the apti-tudes and interests required for successful completion of training.Operational convenience required that the official Army MOS structure befollowed so far as possible consistent with the research findings.

Since the mid 1950's, Army MOS have been grouped into ten occupational

areas based on similar job functions--mechanical maintenance, for example.The aptitude area system was generally tied into the occupational areastructure. The MOS samples in the present research were also grouped byoccupational area, and the mean validity vectors were computed for eacharea. The standard deviation of the validity coefficients for each testin each MOS group was also computed. The results are shown in Table B-iof the Appendix. An alternative grouping of MOS was by Career ManagementFields (CMF), which were being considered for operational implementationin the early 1970's. The CMF are smaller groupings of MOS, each CMF beinga set of closely related jobs. Many of the 39 CMF are expected to requiresimilar aptitudes, knowledge, and interests. Mean validity vectors andstandard deviations of validity coefficients for each CMF for which datawere available were computed (Table B-2).

Each CMF was found t.) be relatively homogeneous, with two notableexceptions--the Field Artillery-Missiles CMF and the Air Defense CMF,both of which included electronics repair and crewman MOS. In thesefields, different tests were found to be valid for the repairmen and thecrewmen. For classification purposed, therefore, the repair and crewmanMOS in these fields were analyzed separately.

In computing the validity vectors, each MOS sample was unit-weighted.In the case of some MOS for which large numbers of men are trained, suchas Infantry, there were samples from several different training installa-tions. Since the sample from each installation was unit-weighted, MOSwith more trainees were in effect multiple weighted. The validity coeffi-cients were not converted to z coefficients prior to averaging. Some meanvectors were based on only a few samples (only two in the Field Cannonand Rocket Artillery compositeY. Others were based on many samples (20in the Electronics Repair composite). Each vectoi was considered to bethe best estimate of the .a'idity of the tests for a given group of MOS.

Comparison of the ineans and standard deviations for occupational areasand career fields indicated t At the CMF were generally more homogeneous. Sincethe expection was that the CP -- or similar groupings--would be used operation-ally for grouping MOS, the dec sionwas made to base tlie aptitude area classi-ficationsystemon the CMF structure. SimitarCMFwere combined, and thecombinations are hereaft-er referred to as MOS groups.

The next step was to simplify the system by reducing the number ofMOS groupings and the number of tests. Test selections were performed onthe mean validity vector f~r each CMF using the forward test selectiontechnique developed by Summerfield and LubinO-'. CMF for which similar testswere selected by this technique were combined. Tests not selected for anyCMF :re dropped. The cycle of combining CMF and dropping tests was re-peated several times.

Before starting with the test selections, three tests that were operi-tional in the old ACB were deleted from the analyfis. The Pattern AnalysisTest was dropped because the items were interdependent, several referriagto the same diagram. The Electronics Information Test was deleted becsusemany items were outdated. The Shop Mechanics Test was deleted because theACB version and the Army Qualification Battery (AQB) were not paralleL.(The AQB is used at Armed Forces Entrance and Examining Stations to r.eter-mine eligibility for service for men in mental Category IV and for enlistedcommitment.) For each of these tests, an experimental test that wa!, equal-ly valid--or more valid--was available and was substituted for the deletedtest in the test selection and evaluation procedure for the new ACE.

The first test selection included 28 tests and 25 CMF. The results areshown in Table B-3. This selection could not be completed on four CMF be-cause the multiple correlation coefficients were greater than unit.yA4

SSummerfield, A. and A. Lubin. A square root method of select-.ng a min-

imum set of variables in multiple regression. Psychometrika, '951, 16271-284.

SThe multiple correl-tion coefficients greater than unity reflects someof the problems in conducting research in an operational setting. Mul-tiple correlation coefficients greater than one are an impossibility ifthe data are complete for all cases in the sample. In the Ary opera-tional setting. however, obtaining complete data on all cases is ex-tremely difficult. Usually, in fact, a large percentage of cases havemissing scores on one or more variables. Ideally, cases with missingdata would be dropped from the sample when computing the multiple cor-relation coefficients. However, the sample size is often barely ade-quate when all cases are included, and to drop cases with missing datawould frequently result in inadequate samples. The researcher is facedwith the dilemma of including all cases, even those with missing data,a step which means that each statistic is based on the maximum numberof cases, or of dropping cases with missing data, a step which satisfiesthe canons of statistical rigor. If all cases are included, the statis-tics may in some cases be inconsistent with one another because theyare based on different individuals and the reduction in numbers is noton a random basis. Dropping cases with missing data has the drawbackthat any statistics oased on a substantially reduced number of caseshave larger standard errors, especially the beta weights computed inmultiple regression.

These CMF were grouped with other CIF in the next test selection on thebasis of similarity of validity vectors. Weights for five tests are re-ported because it was found that this number produced a multiple correla-tion coefficient about equal to that for the full set of tests. Ninetests were dropped after the first set of test selections because theywere relatively unimportant or because they appeared in equations for CMFwhere they were not reasonable.

An expeztation based on the experimental tests was that ElectronicsRepair MOS ( uld be separated from the Electrical-Mechanical MaintenanceMOS. The E 'ctronics Picture Test and Personal Inventory for Electronics(high and lc. level keyr) were designed to make this differentiation.Another expectation was that the Medical MOS could be differentiated fromthe other General Technical MOS such as Intelligence and Topography. TheBio-Chemical Information Test was designed especially to select men forMedical MOS. Results did not support use of these tests to make the de-sired discriminations. The Electronics Picture Test and Personal Inven-tory for Electronics were therefore dropped after the first round of testselections because they had little valid variance. The Bio-Chem Test wasvalid for Medical MOS, but also for Construction, Chemical, Administrative,and Information/Audio-Visual MOS. The Electronics and Electrical Repairand MedLial MOS were kept separate in the hope that with fewer variables,a clearer picture of the uniquely valid tests for these MOS would emerge.The Crewman and Electronics Repairman MOS in Field Artillery-Missiles andAir Defense CMF were separated to determine if separate equations wereappropriate for these two job groups. The net outcome left 19 tests and19 MOS groupings for the second round of test selection.

The first five tests selected in the second round of test selectionsfor each cluster of MOS and their beta weights are shown in Table B-4. Theresults were used to pool clusters that had similar composites of validtests. Infantry-Armor and Combat Engineer MOS were similar. Electronicsand Electrical/Mechanical MOS also had similar composites. Missiles andAir Defense operators, however, were distinctly different from ElectronicsRepairmen. Clerical, Administrative Finance, and Supply MOS were closeenough to suggest rhat they too could be pooled. Mechanical and AircraftMaintenance MOS were also similar, as were Combat Surveillance and Communi-cations Operations. Another grouping that emerged from the data was acombination of Motor Transport, Missile Operators, and Food Service MOS.

A third and final round of test selections was conducted on these newgroupings of MOS to determine whether the MOS could be further combined.The results of the test selection are shown in Table B-5. Two tests, theSubtraction and Division Test and the Science Test, were deleted from thebattery because they had little unique validity. A third test, Tools, wasdropped. Since it appeared for only one CMF--Field Cannon and Rocket Artil-lery--to which only a small number of men are assigned, it was not judgedsufficiently useful to include in the battery. The Field Cannon and RocketArtillery composite would have had higher validity with the Tools Test(r = .69) than without (r = .62), but the increase was deemed worth less

-8-

than the administrative cost of including the test operationally. The MOSgroupings and test composites that emerged from this analysis were consis-tent with prior information about job families and test validity, and nofurther search for different combinations of MOS or tests was considerednecessary.

Nine MOS groups could be differentiated on the basis of the tests in 4the new ACB (Table 2). The new MOS groups are generally similar to thosein the old system. Infantry, previously a separate group, was combinedwith Armor and Combat Engineering to form a Combat group. The positionof Missiles Crewrrzn in the occupational structure has always presenteda problem. In the old system, they were part of a heterogeneous groupcalled AE for Armor, Artillery, and Engineering. In the new system, Armorand Engineering were combined with Infantry, as already noted, and theMissiles Crewmen were found to require the same test composite as MotorVehicle Drivers and Food Services. The latter three MOS were combinedto form the OF group. OF has no counterpart in the prior system. Allthe Electronics and Electrical Repair 11OS were combined in the EL group,even though special efforts were made to separate che more complex elec-tronics repair MOS from the more mechanically oriented electrical repairMOS; the EL groups in the old and new systems are virtually identical.

The SC (Surveillance and Communications) group has no counterpart inthe old system. SC combines radio operator MOS, which formed the old RCgroup, communications center operator, MOS which were in the clericalgroup, anc combat surveillance and target acquisition MOS from the old AEarea. The remaining MOS groups, Mechanical Maintenance (MM), General Main-tenance (GM), Clerical (CL), and Skilled Technical (ST), are similar inthe two systems. Drivers were removed from the old MM, and a few otherminor changes were made. The General Technical (GT) label was changed toST in the new system with a view to reducing the surplus meaning attachingto the GT label which has tended to be associated with IQ.

Validity of Compo!;ites

Each MOS group has associated with it a test composite that is usedas prerequisite for assignment to an MOS in that group. Symbols forMOS group and associated test composite are the same.

The final grouping of the MOS, the tests retained in the new ACB, andthe tests selected for each composite all interact to form the new aptitudearea system. Each aspect of the system exists only in relation to otheraspects. The content of the new ACB, with changes from the old ACB, isshown in Table 3. The tests are grouped in four domains: 1) general abil-ity, which has been expanded from three to five tests; 2) mechanical abil-ity--one test, Shop Mechanics, has been dropped and replaced by TradeInformation; 3) perceptual ability--one test, Army Clerical Speed, hasbeen replaced by Attention-to-Detail; and 4) the Self-Description Inven-tory, to which three new interest scales, Attentiveness, Electronics, and

1 $44o V 0 0)0

o 040 01

(1 -4 W

r- 4 1-4

0 -r ý

(o 4Jd -, ) 0 1." "4 ri W -L .) UCL -4 .,4 () 0 c "-41

:3) N( 41 ) r0l C: 14 "c0 ~4) 0) (4- C/ co 0$4 w4- Q)C) 4-4). 0 Q) Q) 5 1).

4 - -4 u uw 41 -4 w 40) 04) 4J H1 rý 41 UCP E-444 r-4) <

0a Q 1.4 r ,

0 ý4 C -4r4-,

14 ca m0

co( 0,)0

"0 as -4 0 C ) ca '-4iu5 ~ CO3 4 En . (0 > CO

0) p zC 0. 4-i Cr ,4 --C4 0) a) w J410 4 ,

-4 a,) CO -4 r4 (o x -4l-'4 u p -a 01)4r

1-) $4 W C3 ) W O Cm -ý V ~ U) -4 .J -4 (1 d -4 U 410 c-4 4 aC

w r. 5.,4 4-,4 --4 C1 COi r U H0Cc CO rA 04- cn > 0) CO s.n "-r- V r

4 Q) U) 0O) 14 -4 u 0 (a -'-01 "-4 -,4 0), Q M -r 0) V a)

E4 00

4t-

-10

Maintenance, have been added. Each of these fccres was found to be avalid predictor of training success in one or more MOS groups. The va-lidity coefficients of the new ACB tests for each MOS group are shownin Table B-6 (Part 1) and the beta weights are shown in Table B-7 (Part 1).The validity coefficients and beta weights for the tests not included inthe new ACB are also shywn (Part 2 of Tables B-6 and B-7, respectively).

The nine new aptitude area composites are shown in Table 4. All the

composites contain at least one test of general mental ability; ArithmeticReasoning occurs in seven composites, and other tests of general abilityoccur in the remaining two composites. Thus, the new composites have aheavier weighting of general mental ability than the old composites. Thenew composites are also more complex than the old ones. All the old com-

* posites contain two tests, while the new ones contain at least three, andfive of the nine composites contain five tests. The operational signif-icance of the changes in the composites is discussed in the 3ESRL Techni-cal Research Report, TRR 1177, An Improved Differential Army ClassificationSystem•n.

An additional composite, OT (also shown in Table 4), is composed ofArithmetic Reasoning and Word Knowledge. In the old system the GT scoreserved a dual function of selecting persons for the GT MOS group anddetermining which men were qualified to take additional tests such as theOfficer Candidate Test and Flight Aptitude Selection Test. The formerfunction is filled by the ST composite in the new system, while the latter

function is retained for the GT score. Many Army regulations and testingprograms are based on the GT score. Since the GT score is so widely usedand accepted throughout the Army, it was retained at least temporarily inthe new system.

The beta weights for all tests in the new ACB are shown in AppendixTable B-8. The tests with highest weights in the full regression equa-tions were almost always selected for the composites.

The multiple correlation of the new ACB composites with traininggrades is shown in Table 5, along with the multiple correlation coef-ficients of the full composites of 35 variables, which include 11 oper-ational ACB tests, 20 experimental tests, age, and years of education.The loss in validity for the shorter composites is substantial for someMOS groups, especially Field Artillery (FA). One variable that had alarge beta weight in most cases was age, as shown in Table 6 and in TableB-(, which shows the regression equations for all 35 variables. Evidently,the more mature men are often better in training than their test scores

Maier, M. H., and E. F. Fuchs. An Improved Differential Army Classi-fication System. Technical Research Report 1177, Behavior and SystemsResearch Laboratory, Arlington, VA. April 1972.

- I-

00 V0 V

00 c u

,,• o

r- 4) I W1 W1 1- " "0"o • "ca 9) d) 0 60 c0

4o 4J'1 IJ• 4J3 - -0 o J:6 4o CL lu o $4 4- J'

i.-4

000 000 0 0 0::00 0

0O• u - o.u a0'"o v 0.) .-4

1-44

$4100

1-o ,44

C ca• 41 0 )

0 ccao

E) Cl) C4

0 4J 0 4, F-4

w a 44 4- 0 -,4 (1)

- $• CO 4J 4-4 Go .a M C:j

, - 0 Q 0- 0 0 >C 0 '< 0

-12-i

E- k ca- 4-0C) u0 U 4Q c

Cd 0J u-~ ., V 00 *4 0 "'H *'4c -405-4 $i . ~ .4 9:0 C c,o ~ ~ C P.. 41 o 0CO,4Cu - 4. U)w 0- 0).-41. 0 4cov-m

.C04 CD d)-4 0 a04 c

00

o 0' W Q -. o---4- < H-r4 r-f P.. A L -co'Nc )d

'-0) (n( a > 4 L0 --4 .14 01- Q) >.' W- MQ)00 r4 U - 4 i u-C. - 4-i, W C/) m

> C4 0 '-i0 *-4 w 04J0)44M 0 a4)4 4 j c Q)~z0 -44 -ý

_- > 00 -

4j ~ 41 Q,.4 "-I 0C'C4) ̀4 C-i >1 0H -Lc >

a)1 C -4 co ()r C ) 0 -A4J'.w- 4 C11 4 0C :

p440 Q r_ U - 0 5 40 M 3 a 4. J ) r_ - C13C:41 a-0) C C) -, 4 u c a L 0.J~ 4 ol w )1)rM -A W4 r.4 )U4JM- 0 ,Q) < :3 / Q o : 4) U 1J-4 c

-12-

Table 4

NEW APTITUDE AREA COMPOSITES

Test Aptitude Area Composites

General Ability Tests CO FA EL OF SC 1,4 GM C1, ST _T

Arithmetic Reasoning (AR) AR AR AR AR AR AR AR ARGeneral Informat.ion (GI) GI V1Mathematics Knowledge (M•K) MK MK MKWord Knowledge (WK) WK WY WKScience Knowledge (SK) SK SK

Mechanical Ability Tests

Trade Irformation (TT) TI TI TIElectronics Information (El) E1 El ElMechanical Comprehension NO MC MC MCAutomotive Information (AI) Al Al AI

Perceptual Ability

Pattern Analysis (PA) PA PAAttention-to-Detail (AD) AD ADAuditory Perception (AP) AP

Self Description

Combat Scale (CC) CCAttentiveness Scale (CA) CA CA CAElectronics Scale (CE) CEMaintenance Scale (o.) CM

Legend: Aptitude Area Composites

CO=Combat SC=Surveillance and CommunicationsFA=Field Artillery MM=Mechanical MaintenanceEL=Electronics GM=General MaintenanceOF=Operator and Food CL=ClericalGT=General Technical ST=Skilled Technical

GT used only to determine who is qualified to take additional tests such as the OfficerCandidate Test.

- 15-

z~ .....

Uindicate. Age was not included in the new composites because its use in

Si selection and classification has broad iniplications, and a thorough pol-icy review would be required if it were to be used operationally.

Validity of Education, Age, and Selected High Schoci Courses

For some Army training courses, completion of a certain academiccourse is prerequisite. For example, high school algebra is required foradmission to the Field Artillery Operations and Intelligence AssistantCourse. The oresent research provided opportunity to decermine empiri-cally any unique predictive validity associated with completion of se-lected high school courses.

As part of the experimental tetting, examinees were asked to statewhether they had taken certain high school courses: biology, chemistry,general science, electrical shop, hygiene, physics, earth science, algebra,woodworking, and metalworking. The yes-no responses were correlated withfinal course grades in each sample and beta weights computed. Years ofeducation and age, obtained from official records, were also included inthe validity analysis. The regression equations included--in addition toten high 3chool courses, age, and education--scores on the ArithmeticReasoning and Automotive information tests of the ACB and three interestscales from the Classification inventory--ELectronics, Maintenance, andAttentiveness.

The mean beta weights indicated little unique validity for the courses,most coefficients being near zero (Table C-i). Although the validitycoefficients for completion of the courses were positive, the beta weightswere usually small, and about half were negative. Level of education oryears of schooling completed was a valid predictor of training performancein most MOS groups. ln previous researchf', it was found that level ofeducation contributes to the validity of the ACB for selected groups ofMOS. In the present research, level of education was found co be uniquelyvalid w.hen both selected ACB tests and specific courses were included inthe regression equations, as indicated by the high weights ii. Table 6. Thepositive weights mean that men with lower levels of education such as highschool dropouts and younger men do less well in MOS training than theirtest scores indicate, while men with more education such as college grad-uates and older men do better in MOS training than would be expected fromtheir test scores.

Maier, M.H. Effects of educational level on prediction of trainingsuccess with the ACB. Technical Research '•ote 221. Behavior andSystonms Research Laboratory. Arlington, VA. January 1971.

-14-

-4w

>'CM~ .

>cv r- -11 0 -:j 0 n in 00 c .4

'-,4

4-4

0

U)'

0 0

a c

Li") 4. ., 4J

4) a).40 (v') ID cý -11 %D r, cc c a' ID'

E-4 0

-4J

'-4

H- -4

CCu

0

'.o

Cu

"-4-

Ia. pw 0 Cu a') cl) a)

'.1 0 a

(1) CL j -,4 u-,4 a) Cu -4 a) '4)

-4 C W-4"04 C 0 -4 .4 C4

w ).. -4 C o u-

Z r 0 04 -> W. 44Cu

a) Q a ) s-. u Q) .,*~4* ). ~ 0 '.4 -4 C. 1 ) 0 4 .

u .u :74 0 En Q . . '

0 0 ' ) . E'-4

I'15-

Table 6

MEAN BETA WEIGHTS OF LEVEL OF EDUCATION,AGE, AND ARITHMETIC REASONING

(weights abstracted from Table C-1)

MOS Level of ArithmeticGroup Education Age Reasoning

CO (Combat) .01 .OC .20

FA (Field Artillery) .23 .11 , .23

EL (Electronics Repair) .07 .09 .18

OF (Operators and Food) .06 .10 ,03

SC (Surveillance andCommunications) .06 .05 .19

MM (Mechanical Maintenance) .07 :.18 .17

GM (General Maintenance) .10 .16, .21

CL (Clerical) .18 .05 .20

ST (Skilled Technical) .17 .00 .17

I-

S - 6

EVALUATION OF THE NEW APTITUDE AREA SYSTEM

Re duction of Attrition

A major ol-jective in developing a new aptitude area structure was to

reduce attrition from Army training schools. The Taylor-Russell Tables,-'were used to estimate the effect of the new system on failure rates. Theaverage validity coefficient across MOS groups increased from .55 underthe previous battery to .65 under the new. Assuming that 80 percent ofan unselected group would successfully complete the typical Army trainingcourse and that 40 percent of the input is excluded from the typicalcourse, then the failure rate, according to the Taylor-Russell Tables, is

9 percent. The observed failure rate of all Army trainees during calendaryear 1969 was ý.1 percent; the assumptions therefore appear reasonable.With an increase in the average qalidity coefficient to .65, and making

the same assumptions, the expected failure rate would be reduced to 7percent, or by about 20 percent.

The model on which the Taylor-Russell Tables are built assumes a sin-gle predictor and a single criterion. The Army's differential classifi-cation system uses multiple predictors and criteria. The univariate modelis not entirely appropriate, but can serve as a convenient guide to esti-mate effectiveness. A more appropriate model for estimating predictiveeffectiveness ip based on computerized simulation of an input population

and differential classification§-8 Such a model was employed, and theresults supported those found for the simple univariate approximation.

Simulated Evaluation of Performance

In the simulation runs, vectors of normal random deviates were genez-ated and transformed by the ACB population convariance matrix to resemble

scores from the Army enlisted population. Each vector represented theoperational and experimental ACB scores of a randomly selected individual,

plus level of education and age. Aptitude area scores were computed foreach simulated individual or entity, and the entity was then assignedto a job opening. Allocation to job areas was done in such a way as tomaximize the predicted criterion scores summed across all entities andjob areas. Each job area was assigned a quota that equeled recent oper-

ational Army input to the job area. Thirty samples of 500 entities were

Taylor, q.C., and J.J7 Russell. The relationship of validity coefficientsto the practical effectiveness of tests in selection--discussion and

tables. Journal of Applied schology, Iq5Q, 565-578.

-- Niehl, Eli;,abeth and R.C. Sorenson. SIMPO-I entity model for determining the qualitative input of personnel policies. Technical ResearchNote 10. (AD 851 2G',). Behavior and Systems Research Laboratory.

Arlington, VA. January O•.

generated. The quotas are shown in Table D-1.

After the entities were assigned to the MOS groups, their expectedperformance was evaluated by computing predicted criterion scores usingall 55 variables in the prediction equations. The mean predicted perfor-mance score was computed for each MOS group for the entities allocated tothe area. This figure, c[lled the allocation average, shows how muchgain or loss in predicted performance was realized by allocating themen to jobs on the basis of aptitude area composites.

The statistics required to perform the simulation are presented inthe Appendixes. Table A-i presents the intercorrelation matrix of opera-tional and experimental variables used to transform the normal deviates.Table D-1 presents the quotas for each MOF group. Table D-2 presents thebeta weights of the 53 variables for each MOS group in the new system.The beta weights were used to obtain predicted criterion scores for eachentity in its assigned job area.

Mean Levei, of Predicted Performance. The results of the simulationruns are shown in Table 7. In the prior aptitude area system, eightaptitude area scores were computed for ecach entity and the entities wereoptimally allocated to one of eight job areas. Mean predicted criterionscores were obtained in each sample of 500 entities. The same entitieswere also allocated to one of nine job areas on the basis of the newaptitude area composites. Predicted criterion scores in the area of as-signment were also computed from the regression equation of all 55 vari-ables for that area. The mean predicted criterion scores were calculatedfor each job area under each assignment system. An overall r.,an acrossjob areas was also computed. The values shown in Table 7 are the meanpredicted criterion scores under the two aptitude area systems.

The overall level of predicted performance was considerably higherunder the new system, 104.6 for the new versus 102.6 for the previoussystem. If the men were assigned at random, without any prior knowledgeof their skills and aptitudes, the average predicted performance wouldbe 100. When aptitude scores are used in making assignments, the pre-dicted peformance is, of course, increased. The amount of increase isa complex function of several factors, including quotas (percentage ofmen assigned to each MOS group), number of MOS groups, and validity ofthe aptitude area scores-'. The major source of the increase--4.6 pointswith the new system compared to 2.6 with the prior system--is undoubtedlythe higher validity of the new composites, since the number of job areaswas about the same, 9 versus "3, and the quotas in the job areas were ofgenerally the same order.

Brogden, H.E. Efficiency and classification as a function of numberof jobs, percent rejected, and the validity and intercorrelation ofjob performance estimates. Educational and Psychological Measurement,1lV)O, 1"), d 1 - IO.

-18-

Table 7

MEAN PREDICTED CRITERION SCORES UNDER THE NEW APTITUDE AREA SYSTEMAND UNDER THE PREVIOUS SYSTEM

Old Aptitude New AptitudeArea System Area System

MOS MOSGroup Mean Group Mean

IN 100.0 CO 105.1

AE 99.7 FA 106.5

EL 105.9 EL 106.5

GM 98.5 OF 105.5

MM 102.1 SC 107.?

CL 101. 6 MM 105.5

GT 109.0 GM 107.1

RC 112.4 CL 105.6

ST 105.7

Total 102.6 Total 104.6

Two of the MOS groups In the prior system, AE and GM, had mean pre-dicted performance slightly below 100 (Table 7). The entities assignedto these two job groups were below average when all the information avail-able about them contained in the 55 variables was taken into account.Two other areas, GT and RC, were substantially higher than the others.In the new system, all MOS groups had means above average, and all meanswere clustered from 105 to just under 108. Thus, the distribution oftalent was more equitable across all MOS groups tian it has been here-tofore.

Gain from the New Syste . The absolute value of the increase in pre-dicted performance cannot be interpreted directly because there is noscore scale that can be readily applied to the mean values. However, themeans can be interpreted relative tc one another. The increase of 4.6

-19-

points over random assignment attaired with the new system represents a 75percent improvement over the 2.6 points increase under the prior system. Itcan then be said that the utility of the Army personnel classification and

assignment system is increased 75 percent by converting to the new system.

The overall level of performance would be higher under the new system.Since the ACB is used to predict training success, the improved level oftraining performance can serve as a standard by which to evaluate the newACB. In 1970, the median cost of putting a qualified worker in the fieldwas about $6,000, which includes the cost of procurement, basic combattraining, and job training. A reasonable estimate for the combined costof getting a man into the Army and putting him through basic combat train-ing is about $2,000. This leaves about $4,000 as the median cost of pro-viding job training to produce a worker qualified in an MOS. On the aver-age, this approach seems reasonable, although individual cases vary widelyaround the average. The average man performs at a level that is worth asmuch as he costs to train; the below-average performer represents a netloss to the Army, since he performs at a level worth less than the cost oftraining him. The above-average men, following the same argument, isworth more than his training cost, and the Army gains.

One way of measuring the worth of performance is to use the Army stan-dard scale and the training cost as a basis. On the Army standard scorescale, the average level of performance is set at 100. The unsatisfactoryman is defined as one with an expected performance at the level of 80.Expected performance from 80 to 100 covers the range from no net worthto an equal balance between cost and performance. Since 80 representsa total net loss, and the median training cost is $4,000, the man withan expected performance of 80 is a loss of $4,000 to the Army. As thescale of expected performance is ascended, the increased performancebegins to offset the training cost until cost and performance are bal-anced at the average level of 100. Assuming a linear increase, eachpoint increase between 80 and 100 worth t4,000 divided by 20, or 2O00.

The scale also extends to the positive side. Each point ot the in-creased expected performance on the above-average side is also worth$200. With the old ACB, the average expected performance was 2.5 pointsabove what would be realized if the men were assigned on a chance basis,which assumes no knowledge of capability to perform. This gain meansthat the Army has been getting an extra $500 worth of performance (2.5nints gain times $200 for each point) per enlisted man because of theimproved assignments. The n'ew ACB and aptitude area system will addan additic-al 2 points to the average expected performance, or T400 moreper man 'he average above the gain already realized by the old system.At a trt - input rate of 200,000 men per year, the gain of the newACB over L c old assumes rather large proportions. With the new ACB, theincreased worth of training performance is *400 per man; with 200,000 men,the Army each year would be getting T80,000,000 increased worth of pertor-mance from enlisted men during their training assignment.

-2C-

The ý80,000,000 worth of increased performance is a net gain that canbe realized by implementing the new ACB and aptitude area 'ystem. Theincreased performance will not result in an immediate corrasponding re-duction in the Army budget. It does mean, however, that for a fixed num-ber of enlisted men, the overall quality of performance will be higher.As the Army manpower strength is reduced, each position becomes moreimportant and the quality of each man's performance more critical.

Another way of looking at the increased productivity under the newsystem is that the number of superior performers would be increased by15 percent and the number of marginal performers reduced by 20 percent.A superior performer is defined as an individual with predicted criterionperformance of 110 or better on the Army standard score scale, and amarginal performer as one with predicted criterion performance of 90 orbelow. Table 8 presents the number of men expected to be marginal orsuperior performers under the two systems. The numbers are based on aninput of 200,000 men.

The gains in performance expected from the new aptitude area systemwould be realized through improved assignments, and not through higher

Table 8

EXPECTED PERFORMANCE IN OLD AND NEWAPTITUDE AREA SYSTEMS

(Based on input of 200,000 men)

Marginal SuperiorAptitude Area System Performers Perfo.'mers

Old 35,000 57,500

New 27,500 67,500

20% decrease 15% increase

-,

¢I

selection standards. In the simulation runs, exactly the same entitieswere assigned under each aptitude area system and their predicted perfor-mance was evaluated by the same set if variables (that is, all 55 predic-tors,. The new composites are better measures of potential, and theiroperational use will result in fewer errors in assignment.

Weighting Tests in the Aptitude Area Composites. The simulation runswere also used to settle the question of the weights to assign the testsin each composite. From an operational point of view, the most desirableprocedure is the simple addition of all tests in a composite. Maximumvalidityv, however, is obtained when beta weights are used.

Several weighting schemes were evaluated to determine the loss in meanpredicted performance when progressively simpler systems are used. Themost effective system used the maximum amount of valid information abouteach individual in making assignmentsP,. Collecting and utilizing theinformation is expensive, however, and a tradeoff must be reached betweencost of collecting and using information and its incremental utility.Costs and utility at the present state-of-the-art are best evaluated byexpert judgment. Simulation runs can provide data on which to base judg-ments.

The allocation averages for the different weighting schemes are pre-sented in Table (). The most valid, but most cumbersome scheme, was to useall 35 variables in each composite--the 11 operational tests, the Lyu ex-perimental tests, plus age and education. When the entities were assignedand evaluated by the full regression equations, the maximum allocationaverage, 107,r, was obtained. However, such a complex weighting schemewould be too ditticult to be used routinely in the field. One simplifica-tion was to drop 15 tests that did not emerge in the test selections, aswell as age and education--variables which were in effect assigned weightsof zero in the assignment process. The beta weights for the remaining1ý2 tests were used in computing the composite scores. For this weightingscheme, the allocation average was about 105.9, a drop of I.X points belowthe maximum.

The composites were further simplified through test selection to thetest content presented in Table 4. The question of what weights to assignthe component tests remaiaed. Three options were evaluated: i) weightsclosely approxinating the beta weights obtained in test selection; 2) sim-ple weights of 1, 2, or 3, roughly approximating the beta weights; and5: unit weights, for which the size of the beta weights was not considered.The three allocation averages were virtually the same: I04.65, 104.04,and 714.55, respectively. The simplified composite resulted in a drop inthe allocation average from 105.2 for the full beta-weighted composites

Soreson, R.C. Optimal allocation of enlisted men--full regressionequations vs aptitude area scores. Technical Research Note ]F'.(AD ,-i). Behavior and Systems Research Laboratory, Arlington,VA. November 1 4-.

4'

Table 9

MEAN PREDICTED CRITERION SCORES WITH ASSIGNMENTSMADE BY DIFFERENT WEIGHTING SCHEMES

Weighting Scheme Mean Score

Beta weights, 55 variables 107.46

Beta weights, 10 variables New ACB) 15.587

Complex weights, New Composites 104.65

Simple weights, New Composites 104.64

Unit weights, New Composites 104.55

of 15 variables to about 104.E for the unit-weighted composites of 3 to5 tests each. The operational convenience of using a reduced number ofvariables in each composite was judged worth the logs of increased pro-ductivity that would result from the full equations. The more complexcomposites would be more difficult to compute--and also more difficultto interpret. For this reason, their use might be resisted by someoperating personnel. The decision was to propose for operational usethe simplest composites, inasmuch as the procedure did not reduce theallocation average below an acceptable level.

Selection of Men with Marginal Mental AbilIty

An important operational concern about the new composites is theireffect on the selection of marginal men. Marginal men are defined asthose having percentile scores 10 through 50 on the Armed Forces Qual-ification Test (AFQT)0 a test of general mental ability used for mentalscreening. Two factors are used to determine mental qualification ofmarginal men: level of education and number of aptitude area scores ator above 90. Under mental standards of the late 1960's, non-high schoolgraduates who scored 10-15 on the AFQT had to have two aptitude area

scores of 90 or better; nongraduates who scored 16-30 had to have oneaptitude area score of 90 or better, Nongraduates who scored 10 orabove on the AFQT but failed to meet the aptitude area requirements were

-2-

placed in Trainability Limited Category 1-Y. All high school graduateswho scored 10 or above were mentally qualified for Army service. Theaptitude area scores used to determine mental qualifications were basedon the Army Qualification Battery (AQB), a series of short test corre-sponding to tests of the Army Classification Battery and administeredat Armed Forces Entrance and Examining Stations.

The new composites are more difficult than the prior composites inthat more marginal men fail to get qualifying aptitude area scores of90. The new composites have a heavier weight of general mental abilitythan the previous composites, and they contain more tests in each compos-ite, three to five instead of two. These two conditions mean that someinductees who under the old system met the mental standards then in ef-fect would fail to do so under the newAl.

To find what effect the new composites would have on the distributionof aptitude area scores among the men given preinduction examinations,additioaal simulation studies were conducted. Simulation runs were madeseparately for each AFQT decile beginning at 21-30 and going through9]-Q99 and for the AFQT score ranges 10-15 and 16-20. Since the aptitudearea scores used for mental qualification are based on the AQB, the AQBvariance-covariance matrix and AQB composites were used. The AQB vari-ance-covariance is presented in Table D-3. To obtain estimates of howthe proportion of qualified men differs under the old and new systems,both sets of scores were generated for all men. Level of education wasalso included in the simulation runs. Based on other data available inBESRL, the proportion of high school graduates in a representative inputgroup was about .75, which converts to a standard normal deviate of about.7. When the normal deviates were transformed to have means of 100 andstandard deviations of 20, all entities with education scores above 86were called high school graduates, and those with education scores of 86and below were called nongraduates. Separate counts of aptitude areascores at or above 90 were obtained for graduates and nongraduates.

"One reason more men of low general mental ability are unqualifiedunder the new system is that more tests are included in each com-posite. The prior system tended to capitalize on the men's twohighest scores to determine qualification. The rest of the testscores have been ignored. In computing the new composites, these

lower scores are included, and the effect is to decrease thehighest composite scores. Many men would therefore be shiftedfrom the qualified to the unqualified category. A second reasonfor lowered scores is that the Arithmetic Reasoning Test (AR) ispart of most of the new composites. Since AR has one of thehighest coefficients of correlation with AFQT of any of the AQB

tests, men with low AFOT scores would tend to have lower scoreson AR than on other tests.

24-

The population variance-covariance matrix of aptitude area composites,years of education, and AFQT was used to generate scores in each AFQT scorelevel shown in Table D-4. The population matrix was restricted in rangeby using AFQT as the explicit selector and reducing the standard deviationof AFQT from the population value of 25.9 to 2.9 for each decile above 2C,1.4 for the 5-point range 16-20, and 1.7 for the 6-point rar.ge iC-15. Themean for each composite and education in the various AFQT score ranges wasregressed by the formula:

X = X + b (A-9O)

where A = midpoint of AFQT score interval

5C = AFQT population mean

b = regression weight for predicting variable X from AFQT

X = pooulation mean of variable

The same entities were generated for each AFQT score range, except that themean scores changed; the variance-covariance matrices remained identicalexcept for different degrees of restriction on AFQT.

Nongraduates have had to attain qualifying scores on one or two AQBcomposites, not including RC (Radio Code). The Army Radio Code AptitudeTest, called Auditory Perception in the new ACB, has not been part of theArmy Qualification Battery cnd is not expected to be included in the AQBunder the new system. The SC (Surveillance and Communications) compositein the new system and the RC (Radio Code) composite in the prior systemwere omitted from the present simulations, since the radio code test scoreenters into both composites.

The distribution of number of aptitude area scores at or above 90achieved by marginal men is shown in Table 10. From these data, estimatesof the increased number of nongraduates who fail to qualify on the AQBcomposites can be obtained. Some entities that have no aptitude areascores of 90 or better under the new system had one, two, or even threescores of 90 or better under the prior system. The converse was not true:An entity who had no aptitude area score of 90 under the old system usuallyhas no 90 scores under the new system.

The distribution of aptitude area scores presented in Table 10 isbased on AQB composites. The distribution of aptitude area scores of90 or better for the entire AFQT range of 10 and above is shown in Table11. Most men-with AFQT scores above '0 can be expected to qualify on at,least one aptitude area, and usually on more than one, although 14 percentof nongraduates in AFQT range 71-40 do not have any aptitude area scoresof 90. Men with no scores of 90 or better are considered unqualifiedfor almost all Army MOS training courses. As a rule, special consider-ation must be given them in making as;signments.

-K-

-I - - m -b m w r _

£JO~~t-. Jt-4n en 0 4fl

1-.C % - -4 1-&n4 - 3 r4n it-.oc

iC-44 -.

00~ 10C..7" n- 4% ( I I

C:

-44

.0 4J> Cc.J. 4 C 'Jc 4 nCJCl

6~ ~ 0.f 0- vi

to5 a) 1 4

0<i -4 -4 v-IL

tt 4.L)Cn C 4 C1 L -r 4 4.r 4 n in 0c

4 41

0 -4Ci ý D u 'O4 3

r- 4 n TOr-Os c 4Cs C nOsC)' D

4,4 C4QU CC 4(

-4

E- -4C1 c 4 .- 4.-

C ~44. 4.3

I 0 0 0 0z E4 E-4 - +

00 42e. t-c&JC-.0 1 en r- 0 41 Jv--I.4 T

o o 1 < F- C-) 0- -* ý 4 F 0-I -4 -I

[-4 4 F4

1-4 EO

C-4 0JvI -I Ce

H 0

444

a) CO 0 0 v- 4 0U 'v-v i ~ 1-440C

-43 (a.I -. -- 4 4

CC 42

000 4C-3 Cn 0vI'CCO 0 4C4. c-4 cI.cn

0OJO 04ICO

U) 0) 5

F~. 0

I < E

-26-

Co -4 ('4 r O' a%~ -n i ' Un ('n f,

r4 cn LA N~ OD 0% ON

'o0, C- 1' A00 It r 0 0 -.1 0 n nw0 -4 -4 '-4 ý

0>

0.0 -4 -4 -4

0' -4 1S r-. a'% .4 0 Nl -.a N ý-4 0 0

0 00'

0 C~-4 In a's0 N n N ' .40 0 0 0co' 1-4

.0~ 0

00t 0 LA ~~ LAO -5 N "

10,0

00 14 en (71 cn 0 %0 N r- 1 0N 4 It oNr C7' 0m

ca Nl c'n - 0 '0 m' CO (n m' LA Nl0 0 4 -4 -4 '-4 .- I

4.4 4ý' A -4 N 0 LA) N -40cc 0 4 ý-4 -4

V 0 fN as r-I CN 0' %0 cnl 10 0'0 '00 -4-ý4 0

0 4 ý4 0ý4 "4 m 0c

0- 0.04-4 01 CI Cl l '' N.40

u 0 NrC. 4 't- 0 0 -,1 4

.0~~~ .n(a00L$4 01. N0 %0 No cl-4 0000)0.- ýU N

(n 0 C N -T4V

H 0 ox

'0 00 4-4 uO

~Z-. 0

Lo m Dr c4a o o r0 i0a -4 N-SO -5 CO 0' 00 0L

4 -4 Cs4 N 4 -4 4 - 0 0

co ~ ~ L Cl 'Ao N ON It -4l COSN-4 o n

t LA '00 N 4 CO 'AN -4

0l '0 00

ON -5 CO O. 'N 'nC4-4 000 C4' -1 5 N-I

a~ 00

'0 L A ' CO CO-4 0 0

0'01 r I I I

Ln 0 '0 '' 0- -4 '4 4) l

'tn '-4 -4 NC*4 M I it)i '0 , Go 0,

As long as aptitude area scores are used as screening standards, theincreased difficulty of obtaining aptitude area scores of 90 or betterhas implications for personnel procurement. More :nen in mental CategoryIV would need to be examined under the new system to obtain a specifiednumber of qualified men. The severity of the problem depends on otherfactors, such as quotas for men in this range of ability, educationalrequirements, and size of the available manpower pool. On the average,men qualifying under the new system will have higher levels of generalability and can be expected to adapt better to Army life.

CONCLUSICN

In sum, the new Army Classification Battery and aptitude area systemwere found to be superior to the previous system. The average validityof the new aptitude area composites across all MOS is higher than that ofthe previous composites. Through simulation runs, the new measures werefound to result in higher levels of predicted performance to be achievedby identical sample inputs. In the simulations, exactly the same menwere assigned by both the old and the new ACB; the gain could thereforebe realized without keeping more underqualified men out of tne Army.

While the new system makes it slightly more difficult for men in thelower mental categories to achieve the one or moie aptitude area scores of90 or higher requisite to assignment to MOS training, the men who do qual-ify are more likely to experience success in the Army. This result isparticularly important in considering a n-odern volunteer Army, in allprobability much reduced in size. Under those conditions, it would bemore critical that each space be filled by a man wich high likelihood offilling it competently. Accurate assessment of aptitude at time of entrybecomes even more Important and mistakes in overestimating potential morecostly. The new composites provide better qualified men for Army jobsbecause with the more accurate measurement of capabilities job assig~imentscan be better matched to the men's aptitudes and interests.

i

S* -28-

- * , ,<•'

LITERATURE CITED

1. Brogden, H. E. Efficiency of classification as a function of numberof jobs, percent rejected, and the validity and intercorrelation ofjob performance estimates. Educational and Psychological Measurement,1999q, 19, 181-1g0.

2. Maier, M. H. Procedures for assigning grades to failures and turn-backs in Army school courses. Technical Research Note 197. (AD 8355•2'. Behavior and Systems Research Laboratory, Arlington, VA.April 1966.

3. Maier, M. H. Effects of educational level on prediction of Lrainingsuccess with the ACB. Technical Research Note 225. Behavior andSystems Research Laboratory, Arlington, VA. June 1972.

4. Maier, M. H. and E. F. Fuchs. Development of improved aptitude area

composites for enlisted classification. Technical Research Report

1159. (AD 831 268). Behavior and Systems Research Laboratory,Arlington, VA. September 1969.

5. Maier, M. H. and E. F. Fuchs. An Improved Differential Army Classi-fication System. Technical Research Report 1177. Behavior andSystems Research Laboratory, Arlington, VA. April 1972.

6. Niehl, Elizabeth and R. C. Sorenson. SIMPO-I entity model for deter-mining the qualitative input of persoitnel policies. Technical Re-search Note 193. (AD 831 268). Behavior and Systems Research Lab-oratory, Arlington, VA. January 1968.

7. Sorenson, R. C. Optimal allocation of enlisted men--full regressionequations vs aptitude area scores. Technical Research Note 163.(AD 625 224). Behavior and Systemns Research Laboratory, Arlington,VA. November 1965.

8. Summerfiele, A. and A. Lubin. A square root method of selecting aminimum set of variables in multiple regression. Psychometrika,1951, 16, 271-284.

Q. Taylor, H. C. and J. J. Russell. The relationship of validitycoefficients to the practical effectiveness of tests in selectiondiscussion and tables. Journal of Applied Psychology, 1959, ý1,565-578.

-29-

APPENDIXES

Appendix Page

A. Table A-I. Intercorrelations of experimental variablesand operational ACB tests 32

B. Table B-I. Mean and standard deviation of validitycoefficients for occupational, areas 53

B-2. Mean avd standard deviation of validitycaeffialent for Careez Management Fields 34

B-3. Beta weights of test selected 'in first

round of test selectiond 36'

B-4. Beta weights of test selected in secondrou.nd of test selections 37

B-5. Beta weights of test selected in finalr~ound of test selections ,383i

B-6. Mean and standard deviation of validitycoefficients for MOS groups: Part 1, ACBtpsts; Part 2, Non-ACB tests, 39

B-7. Mean aud standard deviation of beta weights

for 33 variables by MOS group 41

B-8. Beta weights for proposed'Army ClassificationiBattery by MOS group 43

C-1. Mean and standard deviation of beta weightsfor high school courses and selected variablesby MOS group 44

P-i. Quotas for old and new aptitude area systemMOS groups 45

D-2. Beta weights for operational and experimentaltests 46

D-3. Intercorrelations of old and new Army Qual-ification Battery aptitude &rea composites 48

D-4. Intercorrelation of new Army ClassificationBattery aptitude area composites 49

Preceding page blank-31-

-S 00 10 M ---- '-'- ODr 01'c

(P 43% - O o 00 mm A a,0ýe":

Ci -- .. 40m

In m 10 m 0 m m a,'O -

0- cý0--:, mO a.'0 I z.r 0 V4

0 D' " T0 00N0' N m U C7N0%04 m 0 NO.4m 0 n0'

4NN .. 'INnN a- 4 '1C4 fl44 0 m m n w

F. C% -- N N N0. ,

W

02 2,

2m --- N40'N00-0202-1

C?.,

I-. - 02004N4 NNNO 4 .to m m4 4 g 400D~ N 44 I 0~~

- g. 0

06 0m- 0 0m1 0N 0m 1 0 4 0 -j Z co

N a%. 0% ON h 040 00I NC N0 NY 1N 0' 1000 NNI0

0.0

0 10 00, a, " 'I

NNm N 1 N4 0ND N4I4'In 0AN Y N D t4 'D

02 o

- ~-Cw0 ol 0 a, w El---=0. o ý(1 E 0-0

-32

z-C

- 6 1

o c Wf~fU - E N g N f

0-0~r4~., C-N Ws0.fls0

C~~ 4 O h C O , 0 - - -

zo 7,

-14-

>00

.a, 0

zu d' u

2 T5

4%-ILI

mm" Mm el 4

ono 9 -- n" m- '

In en 0-S 4.n 0a.S N.-S

MM Mill"'4 M" .4 9F

VIFIVI.04 ~ m 2.2 .1 4 n1VIF4 I44S4(F

N.0.Sn.. 44 -04 0~ 4 m00N0.

omIF -1. n FI4 n a InFFIgES

FI.m4 o am g .0 m.0. a ýIFI00 4a Inc..P..4 m FIFI4mI mFiFii "444AI4. 4f 4AF

""A o44F n nS I 4 4 Sn , 4 I4 lo %0 m VIT

* ~ ~ ~ ~ ~ ~ ~ ~ - ell-I4 0 5 NF N0,N Q I0 nS

4O NN" mm Sn4S nVIS nS S 0 4F.41 4

4 0FF N .0 F4N-00xFSnS n4S 4 n4lIS nV 4S n4S .S

FI0 O V I N 0 FI O -500

4 4n.nOn0n.t £a0~0f~..~o

- 4ii N F 4 41^1^4 4-"-SI NS4 N 4m- NN5m

@Sn "n Wf N n t4 MD "44 Nn (III S 4S 454S 4 0

D "4 .11ý (71 %D* 0;4 N M ON ID Nn Sn 4 D S0 0?

- 1) 01) N.co Ic 0SnN 4a 4 a .-m VI,.Im.a)

N ~ ~ ~ ~ ~ .4n41n N mt0404044N 4.s5nm ~ 0* ~~a.J. 0 0~c I.F n4 .N n000n .. a.sN5 .00

* N SnSn n S FIN 4'FI n F N s4 ." Sn N.S

.0 Sn "M oN MN -. VI3 A INN.

88 E 65. 5S'N N f d d t Fd t; U

M~~~-.- 0 -M- - -D

V% a On F 61

m4 an a m s -t .4C rnP. anft an

m o . c

-aam ww -

-'4 4O wo..n m4- - m-

4 D0-G o=m40a r, 4. m

4 4

-n w. nw 4 w .nmc4r I

m 4 - ON 11 . .4m.4

U .4 .4 C4.4 .4 .

4n C444 4 44n4.

.4UV, .4n m .4 i S

ooCA .. 2nCo0" f.4.4O -

.4.4~~L .444L4..

.4 .4 .4.4 .tJiji411 41 - - - -

(A .4 .4 .04jd 91

* cczz

GO.

0 - 0

S(n I- CL Is

In -. tmN -N

we

u. ui5C A.E

1-4 0-44

I-I N

C.J.4 ()D 4 C-4 -4

1-4 00%

0C14 4 C-4C

0

z w

41 C14 0 4

a 0 14~1 iU-43

W0 U)E )0 00 r4-4 -4 -4 0

an 414

0- %D4 0*0 0N ITC0

cc4 E-4 Nn -4 1

r n %D N -14

-44 COCnc

1 O 4-4 ONO r-r-. en 0IA0 Or4 ( -4 N C-4 N 4 r.-4N N(-

C- -4liý

en0% 0 CN 44N- N 14e 04

CN 00* IL

woo 00 ~L m

-4

A. mA -

0 -4i

%V 00

1-4 CMLAý

a.0-4E

4-4u U w" r-. '#4

I. U) '.0

Z) Cn En

4-4 U) E-4'LNcc

4 3I-4E ccE E4 J 0- C04 IAj C

-4ý 0CU 4

E- /2:-4aU) ad CY tI.n 0 LA c

C U%o 4)

14 4-4 C u.,'- 0 (4 C.

- '.0 4 a -,43 '4 C A r

,4O c (U c X

CU tv

$4 ) CUC 0 .-41-4 0 ~-, C.Li COJ..-4 c: .Cu4) .m V A4ca 4 M W *-,4 4 m.04 U WC> x a) w)- -

S (V (D ) W~ 0 U C: a) -,40 -M 4- CL z L) 4)

-38

0 0.

z In C14-4 C4j -4 r4

-4 r ýN 44 -4.--4 -4 -4 -4.-4

cv*tv -4v- CJ-4.

04 % -4 1

m- m- .-m 1 4 r-.4 .-4 -

U4 ODO CJC r 4a%& f)-D Dr i D00%C1 (/2n-40 cn i'a C14 en 1-4

1-4Ci

1.4 04 OV 040C 0TC4I n0r 1 f %04 04L Li nC 4( l)ý 4ý 4ý

00nc n 1 tI cn 1? -4 1-4 1-4

0 m ZLi 004.4L 4 Dr %r-r NCAr-MO %

cc .~I)0 w nnIc4 E-4 Ln IA( en -u 4 to 0 .4 q-4

"4'.

<Ci' 0'-4 n nt . ,4e 4 ý

-0 UD414E- cn w- 04r r , ncL ?L 1 hC 4O 40 %.

4.4 P44r-0A 4 L e n n ntLn 0ý4 4r

ý4 000 >

1-4i''. r- 4wC 0 4J -.T4-4 "r- 00 -4 C7 CE-4 Lim-t% L 4L r4 ý

"4 04

1- 0 ,41 -4 ~ '

0 l~0

0 0 -- 4 -- 4 -

(12

rz.~ rý m -4c .- 4%00 0 n0 4 Cn0 n-4 00 0

m 0.u nL n% q 4-4r4rýrI-

-39-

to Uý'~ J~ r. O l 10

00 .

N~~c MO~fO~-0 M c4-M.M-

U) &ro ,L %r D r NE M 10..~O~~ CC J4 4 ý. f, DN

00

8JU V an W -Lmm n .4-4. - 444-

.4 4 W

.44 M -MN c -a

0'0 c 2

z w 1-.4 .- mNw .4 .. 4 0-4C

00%D 04O 0O 2 U k 40ý r 4'

COL-- -4C-4-4ý N- C 4 -4-ý4 4Lu>

-4jo

C-4. C410 4 M 0 4 11. 4 .440m

10 44. 0

4j 4

4A'.0OrN --0M'N %-0-1.4-4..4I .

V

go (I a% c4 -D"-It Ia, ý 40 0O s

ca 00 NC40CC .

.4~~~~g Cu--~-4 ' uu-t nj 21

04 ~ I CC

9d tn W 00 L

-( c~0~--.t'0 -404 .r-4f('-*

OM -4Itý4 61.4 .-4 w-0%4a rý- D - -

0 0 9r.4t4 0V-4 0 V-4 r -

0000040000 0 0 -4 0 r4 0 0 -4

Sil ItS D t0r

00000-0000 00v4400-00r'0

r4 44 't0%

ca P4 ODr- a0 C4J 0%CV,4%1 0% r0% a1~ ~0 - nr%0 r- LPd0000040000 0 0 ý-4 0 0 0 C.)

41

-14.~ 0<1 00000-4-00 0O'Jr40 0.-400r-4

0 0 00 C)ý4 - 0 0 0 0 '-I 0 0-40 0 '-4

I4 00

to 0 0 t3.Lt a 0 V-40v-4Cr4

E4-

uC E- 00000000P cc -4 0 -4 040 0 0 -4

*0

41 Lr4C) 0 0 40 4 - 0 v- r40 -ý-f 4 .4-C I 1 14

Ci 0 .0v '4- '

1 ClM 0r " o ttn(, nc c - No

000000000 O0.-40000,-a

0-40000000 00.-4.-400000

r4-4 ON () In ý n M L )

000-.400000 00-4-00000

M Ln 4 r4 Pr-it r40 No LnW

DO 000 0 r4r40 0 0 0 ~0,-440 000O

040-400 ý 000000ýr

00 00000r-000 0000-00.-iO

C-4

4 0 M-n, 4TU) -4NMTm000000000 CO -0 0 r-4 0 0 004 I .I

000oD 7 ~ Q0 0 Cn 4N4JC.4ý4 .,4~Z 0000 0000 W 00~4--000-40

000400010NW0M0 r04~4-,.-4,-4

J.U 0 0 -4,-40000 4 -

'-4 44 r% C0 V)-

Cu000000000 0 0r-4 0 -40-4 r-4 ,-4 4

0) Cu

0)- 2 .r-4 - - 4 ý

00 o0E4~~ 4 Z z 0C)C ~ ~

-42- -i ý47

r-4 Na -r4 t c 0 o

W ~ - r a % r- i N 0 ccIL) 0.- 0 0 0 0 0

0 ON 0% 1 4 C- 4 1 -4 N -14

CO< - 0 0 0 0 0 0 0 0

000 0 04 0 0 0 0

P4 0 OD i) r-4 rý r, C1 4 rý 0o 4 0 0 0

E- 4-4 N 0% M' 00 (n ut %.0 0- -40 0 0

00 C4 mc N l 00 N 00 N f %.00- 0 0 0 -4 r-40

1-- %D 00 ci .O N C14 C14 r-

CO 0 0 0-0 0 0 4 0 -

r/3 4 c

H- NF mc 4l 0 4I c c L4-E- 0 0 0 0 04 -4 04 0

4 U 04 NCLc 4 N N ' - -

H ~0 0 -0 r-4 N 0

Im U

E-n w 0-4 0 0 0 -4 N -4

ý4 00 -4 -I rý -- s1 0

0C.. 0 cl 0. -4 0- r -I

~0 0~ o O ~ 0 0 C

'44 400-ý4

Ln J N %0 0in

0)

00 ~ ON 0n 00 %0 l00 4OD0C4: 0-0-40-14ý40 0 C0-0 -0 0 .- 4

N 0 0 0 0 ý 41 4aýNý 1

0nrlN f00.4'-4.40 nLm 00n0 .- 0 It - in0 -O

000000000 00-4400-400

C

ý4 -4-4Cl 4 r, t % 1400Cl -4 .2

4..40.4 ItC 1 cn N'4m$ 00 0000000000 CO 0 0 0

u 00 0'00 0" Lo 00O 01 0 0-ý4ý40 00 0-4 c

0ý0 N C, IIIý L

.. 4 00 (U 0000000 0 0 0 0ý4ý4 0 0

0-4 (A

Q'iI 4 rn-I C I Cn N 0I )r y ,r ta

4) 00 0 0 0 > -4 0 0 0I, f I 0

V L/) 04 L.1 4-4 4 C' -4 ý4 41 (nL ~- 000-1 0.? P- 0% 010000000000 X .40 -.4 0 000 cC

-,4

000000000 -4 0-4 0-40 00 0

00

00 000000000 0 0000 ý 40 -4' 0 ý

-4 _44:1

C m

0E E2

-4 0

ý-4 w-44<

z ---- --- - - -

APPENDIX D Table D-1

QUOTAS FOR OLD AND NEW APTITUDE AREA SYSTEM MOS GROUPS

Old Aptitude New AptitudeArea System Area System

MOS Group * of Input MOS Group * of Input

IN 21. CO 27

AE 16 RT 08 ,

,L 08 EL 08

GM 05 OF 10

MM 16 Sc 06

CL 15 MM 12

GT 16 GM 05

RC 05 CL 12

ST ,12

-45-

04 0 , 0 -T Ný %0 0'% 0 Ný

0n 0 0D 0 0 0- 0 0

cn N it) 0f 0n Nn 'f 0 NtE- 04 0 4 004 - 0 0 0

P4 00 'A4 Czý %D 0 0 '~l 1-0 0 0 0 -4 0- - 0 -4 0

00 0- 04 N 0 %a cn il -4

0 0 0 0 0 0. 0) 0 0 0

P. 4 4 00 0 4 004 0 0

E-444) -It Fn '0 C1N 4 11 %0E- 0 0 04 0 -4 0 0 0

CA I

-4a14 0 -0 0 HI 0 0 0) 0

E- 0 -4 0 - -Hz~ ID MI0 m %

C) u ~ r 4 N -t 'W

4-4

(nC0- 4- '0 N t 0 -t a -4 0

'- 00 - - 0 0 0 0 0 0 0 0

.0r- NQ -

cn -It %.D Mt M %0 0 00 -t)-40 04 0 0 01 0 0 r-4

4-46

o o 00 0 0 0 -

z ,4 C14 4n OD ON 0 tvo ~ 0 0 r-4 0 0 -4 0

o 0 0 0 0 0 0

en rl L LA Mf %D 00 P, IA

o0 0q ý 0 04 0 0 0% 0n 0

~ 4 N 4

C 0 0 0 1

e 0 0 00 0 0 0 0

00 N - 4 e 0 0' I

< -4 00 '-4 0%D .4 0 v-

r~0 N 00 0 0 0 0 0

~~~ 0 r- '4 4 N N 4C

o 0 0 0 0 0 0

LI) CV4 -4 LA) VN 0f 4 n0- 0 - 0 0- 0 0 0

0 0 0 0 - 0 0

'~~ ~~~~ c 4 I 0 ' 0 I

0)0 0 4 0 -4 0 0

mz 0,0 C104 v-4 N IA ' - '-

0 0

o 0- 04 0 u 04 0 4

ou 4 0 0n 0 0 0

N-47

1I-t TLnM% DD% n1

1-4 D W -r % -wtýW'

%Di I r n1:fý "0- ?wk

m ?k( 1 l . nML In fLIA0

r-TWIW"M,4C 0i M0W0t

O~ 0I %1 0r 4r T- OOC1 Oe40 T %D

z Cz

H

ZD E-4 4)P 0r Dw w ,0I - Dm%

0 4

1) 0 0 D r- 0DL)L -O 000 00 00 %

r- %0 w.. Ch"00"w.7 D"m - t.0 -4% 00 Ar wr Dr I

00

Car4 -4

00H

00H . -C - %0 ýr "r Tm0 1V

E4~ CU-t0 O(S0cl . nL 0r wr I t"%W.

H 01En ý ýD )O -- 4ýD .t 0 0 ý () r-40 (0 1 %-0

- E

0. -0

000 EUI ýr n-ZC ý nrzr 0 M 'T If%,'D ý rýoo 0 rý'T L') c %,

z 0 0 4 -48- r D- t l ,rý4r DC

r OD t- P, '0 C4 t- CO -*4

00 -t 0O 0 t.0 r- (3 fC- OD

--T '-4 LA 0 %D Go ON '-4 0 N

WO C4 S "1 00 0 Go (n 0 t4 0OD 0% OD %0 r OD 00 0 'D 00

z0H- 0 ' CO 00 c' - .T en 0 C 0% c'nU 00 COD r- r- tA r-. 0 CO0 Ln '

r U0

4 'I-CFl,4C CO 0 CO CO 0 r- co -It C00i

0

fa , r-. COD r-. 0 CO LA r- A -

N 4 ' LA) -4 It OD 0 1'-F1ze4' 0 t-. CO r, 00 r'- %OF' D LA '0 1-4 U

00 C 0 Cn %D 0 cAi CV LA) 0 3600 COD 0 F- co m' F' 03 --T CO

en COD It LA 00 C1 -4