The systematic application of pre-established rules (or standards) for assigning numbers (or scores) to the attributes or traits of an individual. Measurement.

The systematic application of pre-established rules (or standards) for assigning numbers (or scores) to the attributes or traits of an individual.

Measurement

What is the basic purpose of a test for personnel selection?

What does a score of 75 represent on a test?

What question do you commonly ask when you get back your score on a test

40 45 55 60 70 75 80 90 100

Test Scores

40 45 55 60 70 75 80 90 100

Test Scores

Positively Skewed

Distribution

Negatively Skewed

Distribution

-4 -3 -2 -1 Mean +1 +2 +3 +4

Central Tendency

a) Mode (most frequent score)

b) Mean (average score; [EX/N])

c) Median (midpoint of scores)

Variability (Spread in scores)

a) Range (lowest to highest score)

b) Standard Deviation

c) Variance

Normal Curve

-4 -3 -2 -1 Mean +1 +2 +3 +4 Test Score

13.59% 34.13% 34.13% 13.59%

0.13% 2.14%2.14% 0.13%

Number of Cases

Z score

T score

CEEB score

Deviation IQ (SD = 15)

Stanine

Percentile

-4 -3 -2 -1 0 +1 +2 +3 +4

10 20 30 40 50 60 70 80 90

200 300 400 500 600 700 800

55 70 85 100 115 130 145

4% 7% 12% 17% 20% 17% 12% 7% 4%

1 2 3 4 5 6 7 8 9

1 5 10 20 30 40 50 60 70 80 90 95 100

Relationships Among Different Types of Test Scores in a Normal Distribution

Power law distributions are typified by unstable means, infinite variance, and a greater proportion of extreme events.

Ernest O’Boyle and Herman Aguinis

Z = Raw Score – Mean/Standard deviation

Standard Score Example

• Limit collection of categorical data

Age

0 - 1819 – 2526 – 3536 – 4546 – 5556 – 6585 & Above

Income

0 ------ 10,00010,001 – 25,000 25,001 – 35,00035,001 – 50,00050,001 – 75,00075,001 – 100,000100,000 & Above

Age in Years: _______

Income: ____________

~ I-O Research ~Measurement

~ I-O Research ~Measurement (cont.)

Yes _____

No __________ _____ _____ _____ _____

1 2 3 4 5 Highly HighlyDisagree Agree

• Limit collection of dichotomous data

~ I-O Research ~Measurement (cont.)

• Restrict possibility of missing data

1.2.3.4.5.

48 49 50

Scale Questions

Missing

Missing

Computed score for scale or subscales containing questions #5 and #48 will

also be missing

Absolute versus Relative (Comparative) Assessments

Absolute: “How many hours of TV did you watch last year?

“Is this drink sweet?” or “How sweet is this drink?”

Relative: Did you watch TV more hours than you spent reading the local paper?

“Which of these five drinks is the sweetest?”

• Generally, it is easier for people to make relative vs. absolute judgments (more accuracy and consistency exists)

• People rarely make absolute assessments in everyday activities (most choices are basically comparative)

Limitation with relative assessments and the instances when absolute judgments are vital ---

Scales of Measurement

1) Nominal -- Indicates categories, classification (e.g., gender, race, yes/no)

Stats: N of cases (e.g., chi-square), mode

2) Ordinal -- Indicates relative position; greater than, less than (e.g., rank ordering percentiles)

Stats: Median, percentiles, order statistics, non-parametric analyses

3) Interval -- Indicates an absolute judgment on an attribute (equal intervals)

No absolute zero point (a score of 80 is not twice as high as a score of 40)

Stats: Mean, variance, correlation

4) Ratio -- Possesses an absolute zero point (e.g., number of units produced)

All numerical operations can be performed (add, subtract, multiply, divide)

1st

2nd

3rd

Does not indicate how much of an attribute one possesses (e.g., all may be low or all may be high)

Does not indicate how far apart the people are with respect to the attribute Link

~ I-O Research ~

Interesting fact: Substantial amount of I-O studies are non-experimental (about 50%)

Overall Point:

Best for research to be driven by theories and problem-solving approaches not by methodology/statistics

• Much research efforts in I-O focus on rather trivial questions that can be studied with “fancy” techniques

• Bulk of research has limited applied significance

• Safety in work vehicles: A multilevel study linking safety values and individual predictors to work-related driving crashes. • Beyond change management: A multilevel investigation of contextual and personal influences on employees' commitment to change.

• The development of collective efficacy in teams: A multilevel and longitudinal perspective.

Some Recent Articles in the Journal of Applied Psychology

Study Variables

Multi-level analysis (or hierarchical linear modeling; HLM). Allows for the assessment of variance in outcome variables to be investigated at multiple, hierarchical levels. Related analyses include structural equation modeling and latent class modeling

~ I-O Research Trends ~

• Predicting workplace aggression: A meta-analysis.

• The good, the bad, and the unknown about telecommuting: Meta-analysis of psychological mediators and individual consequences.

Some Recent Articles in the Journal of Applied Psychology (cont.)

Meta-analysis: Statistical approach that allows the combination of results from multiple independent studies on a given topic. It allows a better estimate of the true “effect size,” giving more “weight” to larger studies.

~ I-O Research Trends ~

Some Recent Articles in the Journal of Applied Psychology (cont.)

Moderating variable (or 3rd variable): A variable that affects the strength and/or direction of the relationship between two variables.

Mediating variable: Variable that accounts for (explains) the relationship between two variables

Job enrichment strategies Job Satisfaction Age (as moderator)(The relationship may be stronger for older individuals)

Job enrichment strategies Job Satisfaction Growth need strength (as mediator)

(When growth need strength is considered the relationship between job enrichment and satisfaction goes away)

~ I-O Research Trends ~

Data Analysis

Usage: Approximately 10% of papers published in Journal of Applied Psychology employ factor analysis (Structural Equation Modeling; SEM)

✖ Avoid:

Varimax rotationPrinciple components analysisAutomatically keep factors with eigenvalues greater than 1.0

Use:

Iterative principle factors (least squares, or maximum likelihood)Oblique rotation (no assumption of factor independence)

~ I-O Research ~

Factor Analysis ---

✔

~ I-O Research (cont.) ~Suggestions

1) More use of “archival” data (many are of high quality with large sample sizes; e.g., government statistics on unemployment rates)

2) Longitudinal studies (assessment of change over time)

3) Report confidence intervals and effect sizes in addition to significance levels (e.g., p < .01)