Creating a graph following propensity scores - code not working

[Rose Matthews]

Hello all,

I've never had a problem with this code, but this time I'm using it on another dataset and it doesn't seem to be generating the results
AIM: I would like to create a graph showing post-matching results (normally demaracted with a X)
As you can see the X is missing in my graph

Code used:

global treatment volume
global ylist failure
global xlist var1 var2 var3

logit $treatment $xlist 
predict propensity_score
sum propensity_score

disp .2*0.023 //generates calliper

psmatch2 $treatment $xlist, outcome ($ylist) caliper (0.004) noreplace logit 
pstest $xlist, treated($treatment) both 
drop if _weight == . //drops the unmatched

//This creates graph - which doesn't show the matched values.... 
pstest $xlist, treated($treatment) both graph

Can you please let me know what I'm doing wrong?

[Hannah Beilby]

Hi Rose,

Check the other output from the pstest command, if the Pseudo R squared (Ps R2) in the Unmatched and Matched sample is the same or very similar, it could be that the matched sample is actually overlapping, which is why you can't see it in the graph (x's are directly in front/behind the o's).

I am getting similar issues when:
- I use the option mweight() to try to obtain the graph after inverse probability weighting i.e. mweight(ipw_weights)
- I add the treated() option

Would be grateful to anyone who can explain why neither of these seem to work!?

Try running this specification of psmatch2 and see if pstest improves when both of the above options are omitted:

psmatch2 depvar $treatment $xlist, odds logit

It seems like you were missing the outcome variable (depvar in above) from your psmatch2 code, so that might be part of the problem too...

I can't guarantee my solutions are correct, I am quite new to this... take with a grain of salt!

Thanks,
Hannah