Troubleshooting
Problem
I have used the SPSS RELIABILITY procedure to perform an item analysis. I used three /SCALE subcommands to assign the 30 items in my variable list to three scales of 10 items each, as follows: RELIABILITY /VARIABLES=itema1 to itema10 itemb1 to itemb10 itemc1 to itemc10 /FORMAT=NOLABELS /SCALE(Ascale)=itema1 to itema10 /SCALE(Bscale)=itemb1 to itemb10 /SCALE(Cscale)=itemc1 to itemc10 /MODEL=ALPHA /STATISTICS=DESCRIPTIVE SCALE CORR /SUMMARY=TOTAL . I have complete data for all cases on the Ascale items, but there is some missing data among the Bscale and Cscale items. The sample size reported in analyses for all three scales equaled the number of cases with complete data on all 30 items. The procedure gives me a separate reliability analysis for each scale. I expected each of these analyses to use all of the cases that had complete data for the scale. Is there a way to request RELIABILITY to delete cases on a scale-wise basis?
Resolving The Problem
No, the RELIABILITY procedure 'deletes' (omits) cases that have missing data on any items in the /VARIABLES subcommand. There are no options to alter this feature. One advantage of this feature is that it allows you to easily analyze several scales with the same cases. If you do want to use scale-wise deletion of missing data, you will need to run a separate RELIABILITY procedure for each scale, listing only the current scale items in the /VARIABLES subcommand.
More recent releases of SPSS do not allow multiple SCALE subcommands, so you have to run each scale with a separate RELIABILITY command. In that case, if you want the LISTWISE deletion to cover all scales, specify all items on the VARIABLES subcommand, then only the ones for a given scale on the SCALE subcommand
Related Information
Historical Number
33701
Was this topic helpful?
Document Information
Modified date:
16 April 2020
UID
swg21475254