Evaluation Procedures Dr. Steve Training & Development INP6325
Dec 14, 2015
Evaluation Procedures
Dr. SteveTraining & Development
INP6325
Perspectives on Training
1. Trainee – What if trainee does well in training, but gets fired or laid off for being ineffective on the job?
2. Trainer – What if trainee develops what s/he thinks is effective training, but supervisors won’t let workers use newly learned techniques/skills?
3. Organization – What if organization spent big money on training program, but employees are still not effective?
Applied Questions
Training specialists interested in:1. Whether criteria (performance, $ savings)
indicate improvement following training2. Whether improvement was a result of training3. Whether training will be equally effective with
different group of trainees4. Whether training will be beneficial to other
organizations.
Research Questions
1. Which of two or more training programs is the best and why?
2. Which type of training works best with which type of worker?
3. What type of organization can benefit most by this type of training?
Types of Training Researcher (Randall)
Negativists – Evaluating training is either impossible or unnecessary Often decisions made based on anecdotal evidence
Positivists – Only rigorous scientific evaluation is worthwhile If there’s no experimental data, it’s a waste of time
Often don’t have resources for full-blown experiment
Frustrates (Activists) – All training should be evaluated, but must recognize that evaluation quality varies depending on constraints.
Training Evaluations
Training Evaluation only as good as the: Question that is asked Appropriateness of the criteria Evaluation design
Training Evaluations: Ethical Issues
Ethical issues of empirical study Use of control group means some may not get
training Eliminates promotion opportunities
Control group may feel slighted Trainees act differently if know in experiment
Hawthorne Effect
Formative vs. Summative Evaluation
Formative Evaluation – evaluate whether training program is operating as planned BEFORE it is implemented Focus on process criteria – how training should be
implemented Summative Evaluation – Does training
produce the expected outcome? Trained vs untrained group comparisons’ Comparative summative evaluation – which of
2 or more training approaches produces greatest benefit
Threats to Validity of Training
Internal Validity – did training make a difference in this situation? Training Validity – Does trainee’s training
performance meet criteria for training program? Transfer Validity – Does trainee’s job
performance meet criteria? External Validity – Can training results be
generalized to other trainees or settings? Intraorganizational Validity – Will training be
effective on new group of trainees? Interorganizational Validity – Can training
program be used successfully at other organizations?
Threats to Internal Validity
1. History – Events that happen between pre- & post- tests have nothing to do with training
Contamination of results (ex: layoffs, accidents, holidays, etc.)
2. Maturation – Biological or psychological effects of time on training
Effects of fatigue or disinterest
3. Testing – pre-test sensitizes trainee to upcoming post-test
Trainee prepares for it, or learn from pretest Ex: GRE score GRE course GRE score improves
4. Instrumentation – questionable reliability of test measure
If fluctuation in scores is normal, can’t attribute change to training
Threats to Internal Validity
5. Statistical Regression – Extreme (hi or lo) pre-test scores will often regress to the mean on post-test
6. Differential Selection of Participants – if allow participants to volunteer, experimental group may be more highly motivated
Random selection, random assign, or matched groups
Threats to Internal Validity
Pretest Posttest
• Experimental• Control
7. Experimental Mortality – differential loss of participants between groups
Trainees who did poorly on pre-test may get discouraged, fired, quit, etc.
Training may coincide with other project deadlines
8. Interactions – differential effects that one threat might have on another
Ex: testing effects different in each group due to selection
Threats to Internal Validity
9. Diffusion or Limitation of Treatments – members of one group share information with the other
10. Compensatory Equalization of Treatments – to avoid perceived differential treatment, some alternative might be given to control group
No longer true control, but 2nd experimental group
Threats to Internal Validity
11. John Henry Effect – competition between experimental and control group increases motivation of control group to work harder
12. Demoralization of Control Groups – passive-aggressive response to not being selected for training
Threats to Internal Validity
1. Reactive Effect of Pre-testing (sensitization)- effects of pre-test lead to increase sensitivity of instruction
Increased attention to material that was seen in pre-test
2. Interaction of Selection & Experimental Treatment (representative sample) – characteristics of the group chosen for evaluation may be different than in future groups
Threats to External Validity
3. Reactive Effects of Experimental Settings – Experimental group knows they’re being observed causing them to react differently from future groups
Hawthorne, Guinea pig, or Pygmalion Effects
4. Multiple-Treatment Interference (carryover)– problem with within subjects experiments
participant is exposed to more than one treatment - previous treatments will affect later ones
Threats to External Validity
Training Evaluation Designs
Pre-Experimental Designs One Shot Case Study:
No control group
X T2
One Group Pre-Test/Post-Test: Before and after comparison, no control group
T1 X T2
Static Group Comparison: X T2
-- T2
True Experimental Designs Pre-test/Post-test Control Group:
Controls for most internal validity threats (except diff treatment)
R T1 X T2
R T1 -- T2
Solomon 4-Group Design: Controls for both internal and external threats
R T1 X T2
R T1 --- T2
R -- X T2
R -- --- T2
Post-Test Only Control: Like static group, but with randomization
R X T2
R T2
Training Evaluation Designs
Quasi-Experimental Designs Time Series
Longitudinal method rules out maturation, testing, stat regression
T1 T2 T3 T4 X T5 T6 T7 T8
Non-Equivalent Control Group Same as pre-test/post-test control, but not random
selectionT1 X T2
T1 --- T2
Training Evaluation Designs
Other Evaluation Methods
ANCOVA – use pre-test as covariate in comparing post-test means Example covary measure of ability prior to training Useful when small r between pre & post testR T1 X T2
R T1 --- T2 Correlations – correlate training performance to on-
the-job performance Doesn’t necessarily mean training had effect
Content Validity – SMEs opinion that KSAs covered in training were those identified in needs assessment