Error when using covariates with the formula "dependent_post - dependent_pre ~ treatment"

[mehmeterdemarslan]

Hi,

I am receiving the error below when I include covariates in the estimation and use the formula "dependent_post - dependent_pre ~ treatment" for diff-in-diff.

Error in model.frame.default(formula = dependent_post - dependent_pre ~ : object is not a matrix

I guess it is related to the part in mwa source code where it handles the estimation formula using the covariates given by the user (lines 434-446 in mwa.R). When I checked, I see that part of the code gives the final formula as:

[1] "dependent_post - dependent_pre ~ treatment + troops + pop_1km + elev_1km + nl_1km + logdist + dependent_post - dependent_pre ~ treatment"

However, it works well when I use the default formula ("dependent_post ~ dependent_pre + treatment") and covariates:

[1] "dependent_post ~ dependent_pre + troops + pop_1km + elev_1km + nl_1km + logdist + treatment"

I guess it is a simple glitch where the former formula input does not split in line 435, then added up again altogether at the end of final formula.

Thanks in advance for your time!

[kdonnay]

Hi Mehmet,

Thank you for flagging this bug. I also assume this issue arises from some error in how the formula you entered is parsed in that segment of the code.

I am still actively maintaining mwa but at reduced capacity. Your issue has been added to the things that I will aim to fix for the next release, most likely later this summer/early fall.

My apologies for the delay in addressing this. I hope in the meantime the "standard" diff-in-diff estimation formula that works without issue is a useful alternative.

Thanks again for flagging this!

[mehmeterdemarslan]

Hi Karsten,

Many thanks for the reply. Yes, standard formula is working just well.

Best,

Mehmet