PSYC 516 Final Guaranteed A+
By nature and by necessity, measures of individual differences are ✔️discriminatory
Differential Validity ✔️distinguishing b/w performance on the selection measure and performance on
the job (i.e. diffs in validity coefficients across grps)
Diff. validity ✔️when the relationship b/w predictor and criterion is different for different groups
- if a test is valid for 2 grps, but more valid for one than the other
Diff. prediction ✔️difference in slopes or intercepts across groups
2 conditions that must exist for diff validity to exist? ✔️1. significant diff b/w coefficients obtained for 2
subgrps (ethnicity or gender)
2. correlations found in 1 or both grps are significantly different from 0
Fairness ✔️treating all ppl alike, justly, equitably; having no adverse impact on any grp of ppl
2 dimensions of fairness ✔️1. distributive
2. procedural
distributive ✔️fairness of the outcomes
procedural ✔️fairness of the procedures used to reach a decision
,types of predictive bias ✔️single-grp validity and diff validity
adverse impact ✔️employment practices that appear neutral but a discrim. effect on a protected grp
4/5ths rule ✔️test to calculate adverse impact
- compares rates of selection for lesser represented classes of individuals against the rate at which the
most represented grp is selected
- the lesser represented grp has a selection rate that is less than 4/5ths or 80% of selection rate for the
most represented grp can be used as evidence of discrim or adverse impact
single-grp validity ✔️a test is valid for 1 grp (usually whites) but is invalid (0) for other grps (typ minority
grps)
test-scoring banding ✔️method for referring candidates for selection, addresses this broader goal of
test fairness, as well as the appropriateness of the test-based conducts or rules underlie decision
making-that is, distributive justice
sliding-band method ✔️instead of just selecting the top score, you create a grp w/in standard error for
the diff of the top score and treat them all equal
differences b/w criterion-referenced bands and predictor-referenced bands ✔️1. criterion-ref bands
produce wider bands
2. criterion-ref bands may not be possible b/c lack of criterion-related validity info
3. criterion-ref bands includes the use of available data (may be imperect)
4. criterion-ref bands include 2 reliability coeff, one fore each variable
,sliding band method ✔️allows band to be moved down a score point when highest score is exhausted,
once the highest person is selected, the band moves down to the next highest score
fixed band method ✔️candidates in lower bands not considered until higher bands have been
exhausted
What differences are typically analyzed through the examination of diff validity? ✔️race, ethnicity,
gender
How do you determine fairness? ✔️HRM decisions make justice decisions in employment decisions
example of diff prediction ✔️if an org is using a cognitive ability test to select for employees in an
engineering position, but it has led to a lower proportion of minorities being hired than the majority
zero validity ✔️if the distribution of the predictor/criterion scatterplot is similar for grp and circular
resources for guiding HRM fairness ✔️1. uniform guidelines on employee selection process (1978)
2. standards for educational and psychological testing (1999)
3. Principles for validations and use of personnel selection procedures (2003)
HRM decisions are based on ✔️individual difference measures
moderator ✔️type of variable that decreases the size of the predictor/criterion relationship depending
on what grp of the variable that person belongs to
positive validity ✔️predictor-criterion relationship is the same for both subgrps and elliptical in shape;
conclude fairness, validity, and legally supported
, zero validity ✔️predictor-criterion relationship is the same for both subgrps but circular in shape,
conclude no diff validity, no point to consider predictor
positive validity but adverse impact ✔️predictor-criterion relationship shows differences per subgrp
and elliptical shape; conclude valid and legal adverse impact but only if criterion necessity proven
positive validity for combined grps, invalid for separate grps ✔️predictor-criterion relationship is high
for entire grp but low or zero for either subgrp and elliptical shape; conclude unfair, invalid, illegal and
discrim.
Strategies for reducing adverse impact ✔️1. improve the recruiting strategy for minorities
2. use cognitive abilities in combination w/ noncognitive predictors
3. use multiple regression and other methods for combing predictors into a composite
4. use measures of specific, as opposed to only general, cognitive abilities
5. use differential weighting for various criterion facets, giving less weight to criterion facets that require
more general cognitive abilities
6. use alt modes of presenting test stimuli
7. enhance face validity
8. implement test-score banding to select among the applicants
meta-analyses ✔️are often used to determine if diff validity might exist for certain variables
By nature and by necessity, measures of individual differences are ✔️discriminatory
Differential Validity ✔️distinguishing b/w performance on the selection measure and performance on
the job (i.e. diffs in validity coefficients across grps)
Diff. validity ✔️when the relationship b/w predictor and criterion is different for different groups
- if a test is valid for 2 grps, but more valid for one than the other
Diff. prediction ✔️difference in slopes or intercepts across groups
2 conditions that must exist for diff validity to exist? ✔️1. significant diff b/w coefficients obtained for 2
subgrps (ethnicity or gender)
2. correlations found in 1 or both grps are significantly different from 0
Fairness ✔️treating all ppl alike, justly, equitably; having no adverse impact on any grp of ppl
2 dimensions of fairness ✔️1. distributive
2. procedural
distributive ✔️fairness of the outcomes
procedural ✔️fairness of the procedures used to reach a decision
,types of predictive bias ✔️single-grp validity and diff validity
adverse impact ✔️employment practices that appear neutral but a discrim. effect on a protected grp
4/5ths rule ✔️test to calculate adverse impact
- compares rates of selection for lesser represented classes of individuals against the rate at which the
most represented grp is selected
- the lesser represented grp has a selection rate that is less than 4/5ths or 80% of selection rate for the
most represented grp can be used as evidence of discrim or adverse impact
single-grp validity ✔️a test is valid for 1 grp (usually whites) but is invalid (0) for other grps (typ minority
grps)
test-scoring banding ✔️method for referring candidates for selection, addresses this broader goal of
test fairness, as well as the appropriateness of the test-based conducts or rules underlie decision
making-that is, distributive justice
sliding-band method ✔️instead of just selecting the top score, you create a grp w/in standard error for
the diff of the top score and treat them all equal
differences b/w criterion-referenced bands and predictor-referenced bands ✔️1. criterion-ref bands
produce wider bands
2. criterion-ref bands may not be possible b/c lack of criterion-related validity info
3. criterion-ref bands includes the use of available data (may be imperect)
4. criterion-ref bands include 2 reliability coeff, one fore each variable
,sliding band method ✔️allows band to be moved down a score point when highest score is exhausted,
once the highest person is selected, the band moves down to the next highest score
fixed band method ✔️candidates in lower bands not considered until higher bands have been
exhausted
What differences are typically analyzed through the examination of diff validity? ✔️race, ethnicity,
gender
How do you determine fairness? ✔️HRM decisions make justice decisions in employment decisions
example of diff prediction ✔️if an org is using a cognitive ability test to select for employees in an
engineering position, but it has led to a lower proportion of minorities being hired than the majority
zero validity ✔️if the distribution of the predictor/criterion scatterplot is similar for grp and circular
resources for guiding HRM fairness ✔️1. uniform guidelines on employee selection process (1978)
2. standards for educational and psychological testing (1999)
3. Principles for validations and use of personnel selection procedures (2003)
HRM decisions are based on ✔️individual difference measures
moderator ✔️type of variable that decreases the size of the predictor/criterion relationship depending
on what grp of the variable that person belongs to
positive validity ✔️predictor-criterion relationship is the same for both subgrps and elliptical in shape;
conclude fairness, validity, and legally supported
, zero validity ✔️predictor-criterion relationship is the same for both subgrps but circular in shape,
conclude no diff validity, no point to consider predictor
positive validity but adverse impact ✔️predictor-criterion relationship shows differences per subgrp
and elliptical shape; conclude valid and legal adverse impact but only if criterion necessity proven
positive validity for combined grps, invalid for separate grps ✔️predictor-criterion relationship is high
for entire grp but low or zero for either subgrp and elliptical shape; conclude unfair, invalid, illegal and
discrim.
Strategies for reducing adverse impact ✔️1. improve the recruiting strategy for minorities
2. use cognitive abilities in combination w/ noncognitive predictors
3. use multiple regression and other methods for combing predictors into a composite
4. use measures of specific, as opposed to only general, cognitive abilities
5. use differential weighting for various criterion facets, giving less weight to criterion facets that require
more general cognitive abilities
6. use alt modes of presenting test stimuli
7. enhance face validity
8. implement test-score banding to select among the applicants
meta-analyses ✔️are often used to determine if diff validity might exist for certain variables
