Checking statistical power after propensity score matching

[Nikita Shukla]

Hello! I have performed propensity score matching on my sample data and have dropped unmatched households. This does raise concerns about a possible loss of statistical power and I want to provide evidence that this isn't the case with my data. Although my data is quite large so this should not be a concern, I would still like to do a test to show that it is not but I am not sure how to go about it. I think I have to use the 'power' command but I don't know how to use it to compare the statistical power before and after matching.

Any help would be much appreciated!

Thank you in advance

[David Radwin]

Loss of statistical power is a good issue to address. Can you just do a power analysis of the whole dataset and a second power analysis of the matched dataset and compare your result of interest (e.g., minimum detectable effect size)? The only thing that changes between the two commands is the sample size(s). Ideally you can then say something like "For the whole dataset, the MDES is 1 percentage point, and on the matched sample, it is 1.2 percentage points."

To determine the number of observations after matching, you can use some variation of count if, such as

count if _support == 1

but depends on which matching program and method you use.