propensity score matching using pscore command..

[Olive Bat]

can someone help to understand the matching ways.. my problem is when i do stratification matching ( for urban area) i am getting blanks in ATT, S.E and T statistics..is this because of missing values

[Olive Bat]

is there no one in forum who can guide me on this

[FernandoRios]

Hi Olivie
There could be many reasons for that, including bad matching, improper data, missing values or that you are not calling the command correctly.
For better advice, you should provide a description of your data, the exact command you use to implement PSM, and show the output you obtain. That may be more effective for explaining and understanding your problem.
HTH

[Olive Bat]

Okay i shall share my attached commands.. please check.. cause i get values for overall matching but not urban sector...i am also sharing the results.. plz reply and clear my doubts

[Olive Bat]

Hi Olivie
There could be many reasons for that, including bad matching, improper data, missing values or that you are not calling the command correctly.
For better advice, you should provide a description of your data, the exact command you use to implement PSM, and show the output you obtain. That may be more effective for explaining and understanding your problem.
HTH

Please check my commands using pscore for psm

[Cong Gu]

For panel data, suppose X1 is the variable of main interest (this variable should be an indicator variable), variables x2 to x4 are your control variables, and Y is the dependent variable, the code should be:

psmatch2 x1 x2 x3 x4, out(Y) logit n(1) caliper(.001) noreplacement (this code might be adjusted according to your specific estimation equation)

For diagnostic check:
Step 1. logit x1 x2 x3 x4 (specific to your estimation equation)
Step 2. logit x1 x2 x3 x4 if _weight==1 (specific to your estimation equation)
Step 3.

local variables x2 x3 x4
foreach var of varlist `variables'{
ttest `var' if _weight==1, by(x1)

Hope it may help.

[Olive Bat]

For panel data, suppose X1 is the variable of main interest (this variable should be an indicator variable), variables x2 to x4 are your control variables, and Y is the dependent variable, the code should be:

psmatch2 x1 x2 x3 x4, out(Y) logit n(1) caliper(.001) noreplacement (this code might be adjusted according to your specific estimation equation)

For diagnostic check:
Step 1. logit x1 x2 x3 x4 (specific to your estimation equation)
Step 2. logit x1 x2 x3 x4 if _weight==1 (specific to your estimation equation)
Step 3.

local variables x2 x3 x4
foreach var of varlist `variables'{
ttest `var' if _weight==1, by(x1)

Hope it may help.

sorry still not clear.. mine is a cross sectional data..