* For further details see http://www.jutze.com/research/2013-schult-et-al-differential-prediction/ * Project: Genderfairness * Purpose of this syntax: data analysis for the "Reanalysen" publication * * written by Johannes Schult, johannes.schult@uni-konstanz.de * last updated 2011-05-24 * * Run t-tests for descriptive statistics for the samples * i.e., compare mean scores for men and women on all variables * Sample 1 GET FILE='myhome\set1V1ask.sav'. DATASET NAME set1 WINDOW=FRONT. T-TEST GROUPS=female(0 1) /VARIABLES=hsgpa fgpa test_verbal test_numeric test_vn /CRITERIA=CI(.95). DATASET CLOSE set1. * Sample 2 GET FILE='myhome\set2V1trapmann.sav'. DATASET NAME set2 WINDOW=FRONT. T-TEST GROUPS=female(0 1) /VARIABLES=hsgpa fgpa test_verbal test_numeric test_vn /CRITERIA=CI(.95). DATASET CLOSE set2. * Sample 3 GET FILE='myhome\set3V1ot.sav'. DATASET NAME set3 WINDOW=FRONT. T-TEST GROUPS=female(0 1) /VARIABLES=hsgpa fgpa test_verbal test_numeric test_vn /CRITERIA=CI(.95). DATASET CLOSE set3. * Effect Size d for all for variables (by sex) * Calculate effect sizes for the t-tests. Not the most pleasant syntax, but it works. * Another solution (using OMS as well, but then pooled SDs from the t-test command) can be found * online at http://www.mathkb.com/Uwe/Forum.aspx/stat-consult/1201/Effect-Size-in-SPSS OMS /SELECT TABLES /DESTINATION FORMAT=SAV OUTFILE='myhome\descriptivettestsV1.sav' VIEWER=NO /IF COMMANDS = ['Descriptives'] subtypes = ['Descriptive Statistics']. * Sample 1 GET FILE='myhome\set1V1ask.sav'. DATASET NAME set1 WINDOW=FRONT. INCLUDE 'myhome\crdescriptivettestsaux.sps'. DATASET CLOSE set1. * Sample 2 GET FILE='myhome\set2V1trapmann.sav'. DATASET NAME set2 WINDOW=FRONT. INCLUDE 'myhome\crdescriptivettestsaux.sps'. DATASET CLOSE set2. * Sample 3 GET FILE='myhome\set3V1ot.sav'. DATASET NAME set3 WINDOW=FRONT. INCLUDE 'myhome\crdescriptivettestsaux.sps'. DATASET CLOSE set3. OMSEND. * Calculate d from sample data in new file and arrange it for the table layout. GET FILE='myhome\descriptivettestsV1.sav'. DATASET NAME temp WINDOW=FRONT. * Calculate d: (mean1 - mean2)/sd. COMPUTE d = (LAG(Mittelwert,2) - Mittelwert) / LAG(Standardabweichung,4). EXECUTE. * Leave only one line per sample/predictor. FILTER OFF. USE ALL. SELECT IF (d>-9999). EXECUTE. * Clean up. DELETE VARIABLES Command_ Subtype_ Label_ Var1 Standardabweichung Mittelwert. * Arrange into table (one sample per line). CASESTOVARS /ID=N /GROUPBY=VARIABLE. * Clean up some more. DELETE VARIABLES N. * Save and exit. SAVE OUTFILE='myhome\descriptivettestsV1.sav' /COMPRESSED. DATASET CLOSE temp.