multiple imputation in SPSS - how to pool the results (Friedman test)

#1
Good morning,

I used Multiple imputation to get estimates for missings in a repeated measures design. I want to know about the sensitivity of a reading test (reading t1 - reading t2 - reading t3) so I calculated Friedman test as the nonparametrical alternative to repeated measures Anova. SPSS does not pool the results of e.g. Friedman test. I know about the "Rubin rules" for pooling results, but as I understood I can't just calculate the mean of chi square estimates or p values. Does anyone know how I get pooled results for the Friedman test?

Thank you
 

spunky

Doesn't actually exist
#2
Does anyone know how I get pooled results for the Friedman test?
you will need to use the variation for the Friedman test called the Skillings-Mack test to accommodate both non-normality and missing data.

this test is not available in SPSS. you will need to use something like R to make it work.

and you are correct, you cannot just 'average' test statistics according to Rubi's rules. multiply-imputed datasets have added uncertainty due to the missing data so the variability of the test statistic needs to be accounted for and that is not straight-forward to calculate by hand.