Download presentation
Presentation is loading. Please wait.
1
DM’ing with Multiple Predictors
Advantages More comprehensive assessment Bias suppression Better prediction Greater face validity and Challenges Imposing information processing demands Much higher technical proficiency Greater cost Time lapse expansion and attrition during cycle
2
Strategies for Selection Decision Making
Methods of collecting predictor information Tests, interviews, work samples, application blanks, simulations (e.g., leader, group, customer) Methods of combining predictor information Adding, weighting, formula or judgement based approaches to estimate a composite score Strategies for facilitating decisions Applying a decision scheme that considers available information to produce a conclusive result
3
What are the best methods to collect information?
Mechanical Judgmental No use of human judgment in collecting applicant information Standardized administration of a job knowledge test, cognitive ability test, or personality inventory. Use of human judgment in collecting applicant information Unstructured employment interview, observer rated performance in a simulation, or dinner at Casa Rustica.
4
How should scores be combined to give applicants an overall score (a composite).
Mechanical Judgmental No use of human judgment in combining applicant information Entering applicant scores on interviews or tests into a pre-defined equation designed to predict job performance. Use of human judgment in combining applicant information Reviewing applicant data and applying clinical human judgment to form an impression of applicant promise.
5
Sparkl* Housekeeper Selection Procedure and Application Score Data
Characteristic Selection Procedure Employer Referral Reading Test Agreeableness Test HR Interview Management Interview Maximum score 15 30 65 25 Regression Weight 1 .8 .9 .7 .5 Cutoff Score 7 22 50 6 10 Applicant Applicant Selection Procedure Scores Amanda 55 Dave 14 21 63 11 Carlo 9 29 60 8 12 Becky 5 24 Cindy 23 13 Regression equation: Y = x1 + .8x2 + .9x3 + .7x4 + .5x5 (intercept omitted). The higher the applicant score, the better on the predictor.
6
Multiple regression Compensatory linear composite method Assumes
additive, linear function among predictors-criterion FYI: can accommodate non-linear functions compensatory relationship between predictors Advantages easily applied; robust to violations; allows ranking Disadvantages compensatory model may not be appropriate unstable parameters with small samples requires all applicants to be measured on all predictors
7
Sparkl* Housekeeper Selection Procedure and Application Score Data
Characteristic Selection Procedure Employer Referral Reading Test Agreeableness Test HR Interview Management Interview Maximum score 15 30 65 25 Regression Weight 1 .8 .9 .7 .5 Cutoff Score 7 22 50 6 10 Applicant Applicant Selection Procedure Scores Amanda 55 Dave 14 21 63 11 Carlo 9 29 60 8 12 Becky 5 24 Cindy 23 13 1 2 3 4 5 Regression equation: Y = x1 + .8x2 + .9x3 + .7x4 + .5x5 (intercept omitted). The higher the applicant score, the better on the predictor.
8
Multiple Cutoffs Minimum threshold method Assumes
a non-compensatory, nonlinear relationship between predictors and criterion deficiency on any one predictor is sufficient to justify elimination Advantages intuitive, simple Disadvantages only identifies the minimally competent requires all applicants be measured on all predictors
9
Sparkl* Housekeeper Selection Procedure and Application Score Data
Characteristic Selection Procedure Employer Referral Reading Test Agreeableness Test HR Interview Management Interview Maximum score 15 30 65 25 Regression Weight 1 .8 .9 .7 .5 Cutoff Score 7 22 50 6 10 Applicant Applicant Selection Procedure Scores Amanda 55 Dave 14 21 63 11 Carlo 9 29 60 8 12 Becky 5 24 Cindy 23 13 Regression equation: Y = x1 + .8x2 + .9x3 + .7x4 + .5x5 (intercept omitted). The higher the applicant score, the better on the predictor.
10
Multiple Hurdles Sequential minimum threshold method Assumes
a non-compensatory, nonlinear relationship between predictors and criterion deficiency on any one predictor is sufficient to justify elimination Advantages intuitive, simple, cost effective Disadvantages only identifies the minimally competent validation difficulties bc of missing data on predictors time lapse from search to hire is long
11
Sparkl* Housekeeper Selection Procedure and Application Score Data
Characteristic Selection Procedure Employer Referral Reading Test Agreeableness Test HR Interview Management Interview Maximum score 15 30 65 25 Regression Weight 1 .8 .9 .7 .5 Cutoff Score 7 22 50 6 10 Applicant Applicant Selection Procedure Scores Amanda 55 Dave 14 21 63 11 Carlo 9 29 60 8 12 Becky 5 24 Cindy 23 13 Sequence First Second Third Fourth Fifth Regression equation: Y = x1 + .8x2 + .9x3 + .7x4 + .5x5 (intercept omitted). The higher the applicant score, the better on the predictor.
12
Combination Method Compensatory minimum threshold method Assumes
a “hybrid” relationship exists between subsets of predictors and criterion Initial screening on base of minimum score on all predictors subsequent selection based on linear composite score Advantages intuitive, simple, allows ranking of applicants Disadvantages requires all applicants to complete all predictors more costly than multiple hurdle approach
13
Sparkl* Housekeeper Selection Procedure and Application Score Data
Characteristic Selection Procedure Employer Referral Reading Test Agreeableness Test HR Interview Management Interview Maximum score 15 30 65 25 Regression Weight 1 .8 .9 .7 .5 Cutoff Score 7 22 50 6 10 Applicant Applicant Selection Procedure Scores Amanda 55 Dave 14 21 63 11 Carlo 9 29 60 8 12 Becky 5 24 Cindy 23 13 1 2 Regression equation: Y = x1 + .8x2 + .9x3 + .7x4 + .5x5 (intercept omitted). The higher the applicant score, the better on the predictor.
14
Profile Matching Prototype matching method
Assumes non-linear profile among existing employees generalizes to applicants Profile of Ringers significantly different from that of non-Ringers correlation and squared difference (D2) methods Advantages cutoffs set with deviation from ideal; allows ranking based on proximity Disadvantages requires all applicants to complete all predictors assumes one best profile exists; precludes others validity of separate predictors is overlooked
15
Sparkl* Housekeeper Selection Procedure and Application Score Data
Characteristic Selection Procedure Employer Referral Reading Test Agreeableness Test HR Interview Management Interview Maximum score 15 30 65 25 Regression Weight 1 .8 .9 .7 .5 Cutoff Score 7 22 50 6 10 Applicant Applicant Selection Procedure Scores Amanda 55 Dave 14 21 63 11 Carlo 9 29 60 8 12 Becky 5 24 Cindy 23 13 Regression equation: Y = x1 + .8x2 + .9x3 + .7x4 + .5x5 (intercept omitted). The higher the applicant score, the better on the predictor.
16
Criterion-related Evidence for S*EEP Measures
4. Cut scores must be rationally or empirically established
17
Test Fairness of S*EEP Reading Test Scores
Males r=.32 Ratings Common r=.25 Females r=.19 Reading Test Scores 5. Test scores must demonstrate fairness in application
18
Graphing of Utility Not hired mean Y Hired mean Y
Applicant mean on criterion Cut score
19
More on Utility Estimation
Other Issues Salary info biases judges 40% rule on capitalization context of work greatly affects estimates negative utility scenarios Can be applied to a host of organizational interventions smoking cessation programs Self-efficacy training programs in organizations Sadacca & Campbell (1985) utility of military MOSs War vs. Peace scenario Negative utility of privates in peace time What do dollars mean in war? UNITS.. Military simulations in desert High (1 SD) vs. normal IQ soldiers High IQ unit killed twice as many enemy soldiers Destroyed 3 times as many tanks
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.