30 If an interaction term would be required this would have to be computed.
31 The value of the standard error of estimate of the fair multiple regression equation is entered here.
77. Enter name of the variable that would represent the conditional probability of failure [RISK] in the Target Variable window and enter the CDFNORM(Zyk)32 in the Numerical Expression window.
78. Click Paste, click OK.
79. Highlight the appropriate syntax and execute/run.
80. Click Analyze, Reports, Case Summaries.
81. Transfer the predictor variables [SATIK, PAB, BMV], the predicted criterion performance conditional on the predictor score [PRED_Y] and the probability of failure conditional on the predictor score [RISK] into the Variable(s) window.
82. Click Paste, click OK, highlight the appropriate syntax and execute/run.
83. Save syntax file [to be used on a newly created data file of predictor scores obtained for the next cohort of applicants] under an appropriate name that reflects the battery involved, save the ouput file under an appropriate name that reflects the position and selection date, save data file under an appropriate name that reflects the position and selection date.
32 Click Transform, Compute Variable, select CDF & NoncentralCDF under Function Group, select cdfnorm under Functions and Special Variables. Transfer into Numeric Expression Window. Enter name of standardized criterion cut off [Zyk].
EVALUATION OF THE UTILITY OF THE FAIR PREDICTION/INFERENCE-RULE
[assuming a battery of predictors]
84. Open SPSS33
84. Click: File, Open, Data TELKOM CREPID.SAV 85. Click Data, Aggregate.
85.1 Transfer cc to the Break Variable(s) window
85.2 Transfer the time/frequency and importance variables to the Summaries of Variable(s) window 85.3 Under Save click Write a new data set containing only the aggregated variables (accept the
suggested file name Aggr.Sav)
85.4 Click Paste, highlight the appropriate syntax and execute/run.
86. Click: File, Open, Data AGGR.SAV
87. Click Transform, click Compute Variable Enter name of the variable that would represent the weight for the first performance dimension [W1] in the Target Variable window and enter tf1_mean*i1_mean in the Numerical Expression window. Click Paste.
88. Repeat the foregoing transform-compute procedure to compute the weight variables for all the performance dimensions [W1-W4]. Click Paste, highlight the appropriate syntax and execute/run.
89. Click Transform click Compute Variable Enter name of the variable that would represent the relative weight for the first performance dimension [RW1] in the Target Variable window and enter W1/(Sum(W1,W2,W3,W4))34 in the Numerical Expression window. Click Paste.
90. Repeat foregoing transform-compute procedure to compute the relative weight variables for all the performance dimensions [RW1-RW4]. Click Paste, highlight the appropriate syntax and execute/run.
91. Click Transform, click Compute Variable Enter name of the variable that would represent the sum of the relative weights for the performance dimensions [CHECK1] in the Target Variable window and enter SUM(RW1,RW2,RW33,RW4) in the Numerical Expression window. Click Paste, highlight the appropriate syntax and execute/run35.
92. Click Transform, click Compute Variable Enter name of the variable that would represent the Rc value of average performance on the first performance dimension [RC_RW1] in the Target Variable window and enter 150500*RW1 in the Numerical Expression window. Click Paste.
93. Repeat foregoing transform-compute procedure to compute the Rc value of average performance on each of the performance dimensions [RC_RW1-RC_RW4]. Click Paste, highlight the appropriate syntax and execute/run.
94. Click Transform click Compute Variable Enter name of the variable that would represent the sum of the Rc value of average performance on each of the performance dimensions [CHECK2] in the
33 The procedure described here is not affected by the nature of the prediction model. The validity coefficient entered into the Brogden-Cronbach-Gleser utility equation will be affected though. The procedure described here is therefore identical to the procedure described earlier [but for step 107].
34 Make sure the number of brackets facing left and the number of brackets facing right are the same.
35 Check1 should equal 1.
Target Variable window and enter SUM(RC_RW1,RC_RW2,RC_RW33,RC_RW4) in the Numerical Expression window. Click Paste, highlight the appropriate syntax and execute/run36.
95 Click Analyze, click Reports, click Case Summaries
95.1 Transfer all the variables but cc into the variables window on the right.
95.2 Click Paste, highlight the appropriate syntax and execute/run 96. Save AGGR.SAV as TELKOM CREPID CALCULATIONS.SAV.
97. Click: File, Open, Data TELKOM 2008.SAV
98. Click Transform, click Compute Variable Enter RC_RW1 in the Target Variable window and enter the value obtained in TELKOM CREPID CALCULATIONS SAV [27309,66] in the Numerical Expression window. Click Paste.
99. Repeat foregoing transform-compute procedure to compute the Rc value of average performance on each of the performance dimensions [RC_RW1-RC_RW4] in the validation study data set. Click Paste, highlight the appropriate syntax and execute/run.
100. Click Transform, click Compute Variable Enter the name of the variable that would represent the actual performance achieved by each case on the first performance dimension expressed on a 0-2 scale [R_CRY1] in the Target Variable window and transform the original CRY rating on a 0-200 point scale to a 0-2 point scale by entering CRY1/100 in the Numerical Expression window. Click Paste.
101. Repeat foregoing transform-compute procedure to transform the original CRY rating on a 0-200 point scale to a 0-2 point scale on each of the performance dimensions [R_CRY1-R_CRY4] in the validation study data set. Click Paste, highlight the appropriate syntax and execute/run.
102. Click Transform, click Compute Variable Enter the name of the variable that would represent the Rc value of the actual performance achieved by each case on the first performance dimension [RC_CRY1] in the Target Variable window and enter RC_RW1*R_CRY1 in the Numerical Expression window. Click Paste.
103. Repeat foregoing transform-compute procedure to calculate the monetary worth of the actual performance achieved on each of the performance dimensions [RC_CRY1-RC_CRY4] in the validation study data set. Click Paste, highlight the appropriate syntax and execute/run.
104. Click Transform, click Compute Variable Enter the name of the variable that would represent the Rc value of the actual performance achieved by each case on the composite criterion [TOTWORTH]
in the Target Variable window and enter SUM(RC_CRY1,RC_CRY2,RC_CRY3,RC_CRY4) in the Numerical Expression window. Click Paste, highlight the appropriate syntax and execute/run.
105. Click Analyze, click Descriptive Statistics, click Frequencies.
114.1 Transfer TOTWORTH into the Variables window on the right.
114.2 De-tick Display frequency tables
114.3 Click Statistics, tick Standard Deviation, Variance, Minimum, Maximum, Mean, Median, Mode, Skewness, Kurtosis.
36 Check2 should equal average salary
114.4 Click Continue, click Paste highlight the appropriate syntax and execute/run.
106. Save the data, syntax and output files.
107. Use this standard deviation, along with the correlation between the criterion and the fair inferences derived from the predictor battery calculated in the fairness analysis, to calculate the utility of the once off fair use of the predictor to select a single cohort of selectees from an applicant group by means of the appropriate Brogden-Cronbach-Gleser utility formula.