a week ago
I am doing data analyses using a continuous independent variable, a continuous outcome, and several covariates. I performed multiple imputations on my independent variable of interest (exposure) and am having trouble pooling variances across different imputation datasets for the type of analyses I am doing. Can anyone please help? The two types of procedures I am using are below.
1) Partial (adjusted) Spearman correlations
Here is the code:
proc corr data=final_imp spearman;
partial age bmi;
My main question is: what code can I use in the proc corr to output the covariance structure (variances) from each of the imputed datasets and how do I then combine the results using PROC MIANALYZE? I am essentially interested in a pooled correlation coeff and 95% CI between PBD (outcome) and PCB153 (exposure with multiple imputations) after adjusting for age and BMI (covariates).
2) Adjusted PBD (outcome) means per categories of exposure (PCB153_tertile) using PROC GLM:
proc glm data=final_imp plots=(DIAGNOSTICS RESIDUALS);
model sqrt_PBD= PCB153_tertile age BMI/solution clparm ss3;
lsmeans PCB153_tertile/cl adjust=bon stderr pdiff;
My question is similar here, a) how do I output adjusted PBD means (95% CIs) by the PCB153_tertile (from the lsmeans statement) for each imputed dataset and then pool all of the datasets together using PROC MIANALYZE? I am interested in pooled mean (95% CI) PBD per tertiles of exposure (PCB153_tertile).
I appreciate any input! Thank you!