Missing Data/Imputation Discussion > Mr
Hi Tim,
Great question. You can indeed tell SPSS which variables to impute, by setting a variable's "role" in the "Constraints" tab of the "Impute Missing Data Values" dialogue box. The options you can choose are: "Use as predictor only" (does not impute data for that variable), "Impute Only" (imputes, but does not use the variable as a predictor in other variable's imputation), or none (which allows the variable to be imputed and to be a predictor of other variable's imputation).
However, in your case I don't think I'd use this method. Instead I think your issue is a matter of coding and conceptualization, not of missing values. If you plan to run your analysis with both conditions (and thus both measures) in the same analysis, then they need to be coded to be one variable (e.g. perhaps you could standardize them). Once this is done and there is just one measure of "therapist adherence", then you don't need to worry about missing data in that context. If you plan to run the analysis separately (not sure why necessarily), then you could just select for each condition when running your imputations.
You can get additional info, as needed, in the SPSS manual on page 21 (found at link below):
ftp://public.dhe.ibm.com/software/analytics/spss/documentation/statistics/20.0/en/client/Manuals/IBM_SPSS_Missing_Values.pdf
Jeremy Taylor
Hi,
I wonder if you can help with my struggles with MI in SPSS! In my data set there are some values that are are not applicable for particular cases. These values are not 'missing data', it is that they are not applicable. For example I am using 2 different measures of therapist adherence to protocol depending on which of the 2 therapy conditions the participant was in. So for the adherence measure that is not used I have 'user defined missing values'. However MI also imputes for 'user defined missing values'. Is there a way around this. Is there any way I can set conditions to tell SPSS which data to impute and which to ignore?
Best wishes
Tim