06-21-2017 01:25 PM
I am using PROC MI, PROC LOGISTIC and PROC MIANALYZE for a project where the outcome is dichotomous and most of the exposures are categorical. I am try to make the Table 2 (the unadjusted associations between the exposures and outcome) but when I run PROC MIANALYZE for some of the variables the output does not provide the the 95% confidence limits. I think this is only occuring with the variables that in the original dataset had no missing (so none of the values were imputed). But I still need to report the assoications with confidence intervals. Does anyone know how to get these in the output? Please see the code below. Any advice would be greatly appreciated!
proc logistic data= inj_htn_long; class htn (ref = '0') injury (ref = '0'); model htn (ref='0') = injury time_ctr time_ctr*time_ctr/CovB link = cloglog; by _imputation_; ods output ParameterEstimates=lgsparms; ods output CovB=lgscovb; * create output dataset for parameters, and another for parameter covariances; where _imputation_<=5; run; data lgsparms2; set lgsparms; Variable=cat(strip(Variable),strip(ClassVal0)); run; proc mianalyze parms=lgsparms2 covb(effectvar=stacking)=lgscovb; modeleffects intercept injury1 injury2 injury3 time_ctr time_ctr*time_ctr /* variables you want effects for here */; run;