teffects psmatch in stata

[Gayatri Krishna]

Hi all,
I have a doubt regarding the command teffects psmatch in stata. I am interested in seeing the effect of exposure on health outcome (which is binary) with 13 covariates which are categorical. So my question is do we have to put i. as prefix to all covariates as all of them are factor variables?
i.e which one of the following command should I follow?
teffects psmatch (health outcome) (exposure i.x1 i.x2 i.x3 i.x4 i.x5........i.x13), atet
or
teffects psmatch (health outcome) (exposure x1 x2 x3 x4 x5.........x13), atet
I am getting different results in either case.
In the first case ATT is 0.022 with p value 0.001 and in the second case ATT is 0.016 with p value 0.02.

[Jeff Wooldridge]

Some of the covariates aren’t binary? If they are then there won’t be a difference. The first is correct in general.

To see what’s happening, estimate logit of exposure on the two different ways of including the covariates. That’s what psmatch is doing in the first step.

[Gayatri Krishna]

Thank you for your response!! And yes, some of the covariates are binary and some are categorical or factor variables.
When I estimated logit of exposure on the covariates with prefix i. , I get the coefficients and p-values for the the sub categories as well, but in case of estimating logit without i. just gives the coefficients and p-values for the categories as a whole.
For example, for variable EDUCATION I have the following categories: 1) no schooling, 2) upto primary education, 3) upto secondary education and 4) higher secondary & above. In first case, I get the values for 2) upto primary education, 3) upto secondary education and 4) higher secondary & above keeping 1) no schooling as the base. And in 2nd case, I get the values for education as a whole.