Hello everyone, I have a question about the commands “psmatch2” and “pstest”.
For propensity score matching, the result that I get with Stata version 17 is different from what is shown on a video using Stata version 14.
I summarize the whole process here ( I copied the log file into this rtf file [pstest_question.rtf] ):
-I have a data set of medical records;
-There are 4 variables in the dataset:
age
male
var1
outcomevar
-I generated a matching variable based on var1.
gen matchvar=0
replace matchvar =1 if var1 >14.0
-Then I used psmatch2 and paired the data.
psmatch2 matchvar age male, outcome(outcomevar) caliper(.001) noreplace neighbor(1)logit
gen pair = _id if _treated==0
replace pair = _n1 if _treated==1.
bysort pair: egen paircount = count(pair)
drop if paircount !=2
-Up to this point everything is as it is supposed to be.
-Then I used “pstest“
pstest age male var1 outcomevar
-pstest command without var1 included, works well. But if I include var1 it gives a r(2000) error.
-These are the important pieces of information:
1. var1 (a continuous variable) is the variable that we used for generating the matching variable.
2. In stata version 17, the command < pstest.. var1…> has no output just headings of the table and r(2000) error. [the result is on log file]
3. In stata version 14, the command <pstest …var1…> generate “the variable, mean and bias table” and “the summary of the distribution of the bias table” and a r(2000) error [ I can send a screenshot from the video]
4. command <pstest2 …var1…> works perfectly in stata version 17.
I checked the ado-file of pstest and pstest2, however, I couldn’t find the reason.
I would appreciate it if you could shed some light on it.
For propensity score matching, the result that I get with Stata version 17 is different from what is shown on a video using Stata version 14.
I summarize the whole process here ( I copied the log file into this rtf file [pstest_question.rtf] ):
-I have a data set of medical records;
-There are 4 variables in the dataset:
age
male
var1
outcomevar
-I generated a matching variable based on var1.
gen matchvar=0
replace matchvar =1 if var1 >14.0
-Then I used psmatch2 and paired the data.
psmatch2 matchvar age male, outcome(outcomevar) caliper(.001) noreplace neighbor(1)logit
gen pair = _id if _treated==0
replace pair = _n1 if _treated==1.
bysort pair: egen paircount = count(pair)
drop if paircount !=2
-Up to this point everything is as it is supposed to be.
-Then I used “pstest“
pstest age male var1 outcomevar
-pstest command without var1 included, works well. But if I include var1 it gives a r(2000) error.
-These are the important pieces of information:
1. var1 (a continuous variable) is the variable that we used for generating the matching variable.
2. In stata version 17, the command < pstest.. var1…> has no output just headings of the table and r(2000) error. [the result is on log file]
3. In stata version 14, the command <pstest …var1…> generate “the variable, mean and bias table” and “the summary of the distribution of the bias table” and a r(2000) error [ I can send a screenshot from the video]
4. command <pstest2 …var1…> works perfectly in stata version 17.
I checked the ado-file of pstest and pstest2, however, I couldn’t find the reason.
I would appreciate it if you could shed some light on it.
