Observational Study and controlling variables

[supp]

Hello SAS Community, I am working on an observational study to determine the impact of a program. My approach is to compare the results of the treatment group to a control group. I thought this would be a good approach because I could control for possible confounding variables by using a one to one match on these variables.

After presenting some initial results most of the feedback has been additional variables that should have been controlled for. Is there a way to control for additional variables using the existing control/treatment populations and save my analysis? Or do I just keep collecting more data and constructing new control populations?

For example, if I constructed a control group to compare to the treatment group but failed to control for height of an individual and my populations are not balanced on this variable, is there a way to analyze the data control for height? Such that I can reliably extract the impact of the program?

I have been learning as I go on this project and I suspect this might be a simple/silly question for more experienced statisticians. I am asking in general to figure out what techniques or procedures I need to learn more about.

[PGStats]

I think you should consider using proc PSMATCH for this task.

https://documentation.sas.com/?docsetId=statug&docsetVersion=14.3&docsetTarget=statug_psmatch_overvi... 

[supp]

Thanks @PGStats , while I did not use PSMATCH to create my control/treatment populations I think I got to a similar place using the gmatch macro. Instead of first calculating a propensity score then using that score for the match, I simply matched on my variables. The results looked good. I was able to create a control population that matched on the variables used.

From the PSMATCH documentation, "Provided that the distributions of the variables in the adjusted sample are well balanced between the treated and control groups, the output data set serves as input for a subsequent outcome analysis that incorporates weights or strata or that is based on matched observations. Although the PSMATCH procedure itself does not provide this analysis, many other SAS/STAT procedures can be used for this purpose."

Based on this, if I do in fact have a confounding variable this is not balanced between my populations then I think my best approach is to re-make the control population to balance for additional variables. 

Also in the documentation I see we can perform a sensitivity analysis to help decide if my results should be negated. I will look into this more.

[supp]

To do a sensitivity analysis on matched pairs it seems we need to know the probability of the individual being in the treatment/control group. How do we know this probability? Would we use a logistic regression to model the likelihood of an individual being in the treatment group using the variables used in the matching?