proc mianalyze and non-parametric testing

ablond — Mon, 17 May 2021 06:55:08 GMT

Hi,

I have a question regarding non-parametric analyses on multiple imputated datasets.

I performed the imputation with proc mi, and now I want to use proc npar1way as the data is not normally distributed.

I used the following statement with proc GLM + proc mianalyze as parametric way to compare three groups ; now I want to try this in a non-parametric way (comparing three groups). Is this even possible on an imputed dataset?

thank you for your help!

*statement after proc mi - parametric*
proc rank data = test out = rank_walk3 groups = 3;
var walk_3M; ranks walk_ranks_MI; run;
proc sort ; by _Imputation_; run;

proc glm data=rank_walk3;
ods select ParameterEstimates;
ods select lsmeans;
by _Imputation_;
class walk_ranks_MI (ref = '0');
model delta = walk_ranks_MI dam_baseline/solution;
lsmeans walk_ranks_MI / stderr pdiff cl;
ods output ParameterEstimates=glmparms;
ods output lsmeans = lsmeans_ds;
run;

data glmparms_; set glmparms;
if parameter='walk_ranks_MI 0' then parameter='rank0';
if parameter='walk_ranks_MI 1' then parameter='rank1';
if parameter='walk_ranks_MI 2' then parameter='rank2';
run;

proc mianalyze parms=glmparms_;
modeleffects Intercept rank1 rank2 DAM_baseline ;
ods select ParameterEstimates;
run;

topic proc mianalyze and non-parametric testing in Statistical Procedures

proc mianalyze and non-parametric testing