** ASSIGNMENT 4: CONSTRUCTION OF A MULTIPLE INDICATOR INDEX **. GET FILE='U:\)Datalib\XNAT\ISSP2009\ZA5400_v2-0-0[1].sav'. ** Step 1: inspecting the variables ** ** it is a good idea to look at all the frequencie before you start ** freq v32 to v37, v40 to v43. ** THE FOLLOWING TABLE IS THE BASIS FOR DESCRIBING ALL THE DATA **. desc v32 to v37 v40 to v43. freq v5. ** SELECTING COUNTRIES GOES LIKE THIS ** ** Recode V5 (xxx yyy zzz=1)(else=0) into mycountries. ** Select if (mycountries eq 1). ** I DO ALL THE COUNTRIES AT THE SAME TIME. ** the following procedure standardized all variable to the same metric between countries **. sort cases by v5. split file by v5. desc v32 to v37 v40 to V43 /save. split file off. ** before you split, you have to sort **. ** the desc /save procedure generates Z-scores of the variables **. corr Zv32 to Zv37 Zv40 to ZV43. ** it is easy to see that two variable have to be reverse coded ** ** Z-scores are conveniently reverse coded **. comp Zv35=-ZV35. comp Zv37=-ZV37. ** all variables have now been coded from low: egalitarian high: inegalitarian **. ** Step 2: dimensional analysis **. factor var=Zv32 to Zv37 Zv40 to ZV43 /print=uni corr rotation repr /plot=eigen /criteria=factors(2) /extraction=paf /rotation=oblimin. ** conclusion: there are two dimensions underlying these indicators **. reli /var=ZV32 to ZV37 /summ=total. reli /var=ZV32 Zv33 Zv34 Zv36 zV37 /summ=total. ** there is some evidence that Z35 does not contribute to measurement quality reli /var=Zv40 to Zv43 /summ=total. ** STEP 4: Constructing the index variables **. ** by taking available means **. comp redist=mean(ZV32,Zv33,Zv34,Zv36,zV37). comp redistx=mean(ZV32 to ZV37). comp conflict=mean(Zv40 to Zv43). ** or by calculating component scores **. factor /var=Zv40 to Zv43 /save=reg(1). factor /var=ZV32 Zv33 Zv34 Zv36 zV37 /missing=meansub /criteria=factor(1) /save=reg(1). corr redist redistx fac1_5 conflict. ** STEP 5: MAKE SURE THAT YOU KNOW THE METRIC OF THE RESULT **. DESC redist redistx conflict. ** Making a level of education indicator index **. freq degree educyrs ar_degr to za_degr. recode educyrs (30 thru hi=sysmiss)(97=0)(1 thru 25=copy) into Eddur. var labels Eddur "Duration of Education". means eddur by degree. comp eductp=mean(ar_degr to za_degr). freq eductp. means eddur by v5 by eductp. unianova eddur by eductp /design = eductp /save=pred(edtpdur). desc edtpdur. aggregate /break=eductp /edtpdr2=mean(eddur). desc edtpdur edtpdr2. split file by v5. corr degree eddur edtpdr2. split file off. split file by v5. rank degree eddur edtpdr2 /percent. split file off. desc pdegree peddur pedtpdur2. comp edlevel=mean(pedtpdr2,peddur). corr edlevel redist redistx conflict. ** VALIDATION ANALYSIS ** corr pdegree peddur pedtpdr2 edlevel WITH redist redistx conflict.