How do I fix small Proc Causaltrt ATE estimation?

SASnovic3 — Mon, 18 Jan 2021 23:04:15 GMT

Hello!

I am trying to use the PROC CAUSALTRT statement to find the ATE on SAS version 9.4. The control group and the treatment group are quite different so there are many confounders that I am matching on.

Below is my code. &covariates includes all categorical covariates (about 40). "treat" is the treatment group variable; admission_morb is the outcome variable.

The ATE is estimated at 0.008 with a VERY small confidence interval. Yet a simple logistic regression gives an OR = 0.62 (0.42 - 0.85). I think I am doing something wrong with the PROC CAUSALTRT statement do not know what. Any ideas or anyone else gettin weird results??

proc causaltrt data=propensity_TEST1 covdiffps;
class &covariates /desc;
psmodel treat = &covid2 elx_score /plots=(PSDist pscovden(effects(elx_score)) );
model admission_morb;
run;

Re: How do I fix small Proc Causaltrt ATE estimation?

ChrisNZ — Wed, 20 Jan 2021 00:22:32 GMT

Moved to Statistics forum.

topic Re: How do I fix small Proc Causaltrt ATE estimation? in Statistical Procedures

How do I fix small Proc Causaltrt ATE estimation?

Re: How do I fix small Proc Causaltrt ATE estimation?