Common Support Graphs in Kmatch (PSM)

[Felix Bittmann]

Dear all,
I work with Ben Jann's kmatch ado and got a question concerning the common support graphs. Basically I do not understand what his graph is supposed to show in contrast to the ones I know from other ados or researchers. Consider the testing code below:

ssc install kmatch
ssc install moremata
ssc install kdens

version 15
sysuse nlsw88, clear
kmatch ps union c.ttl_exp i.south c.grade (wage), gen(diag)

kmatch cdensity, name(regular, replace)        //regular Command in kmatch
twoway (kdensity _KM_ps if diag == 0) (kdensity _KM_ps if diag == 1), name(myown, replace)    //own version
graph combine regular myown

I am used to common support diagnostics like this here: https://www.statalist.org/forums/fil...43&type=medium
You see there are two groups, treatment and control. I do not get why in kmatch the groups are called total, unmatched and matched. I create my own version called "myown". You see, this clearly deviates from "regular". Basically, I do not understand how "regular" is a common support graph and how it should be interpreted.

[Felix Bittmann]

I contacted the author and can tell you some news. If you want the version I produced on my own you can use the command kmatch density.
The version that is produced by kmatch cdensity is used to diagnose how the estimand changes when dropping cases that cannot be matched. When cases are not matched, they are removed from the analysis and therefore not used for calculating the ATE. The graphs produced show how this influences your results. If the Total and Matched graph are very close to each other you can be sure that the bias due to the changed sample is not too severe.