Error: Class of the `.Dtreat` variable is class is not supported.

[romkhun]

Hi Grant,

Thanks for making this package. I would like to use this package on my Ubuntu workstation, but it gives me the following error.

I have tried reinstalling R and this package, but the same error occurs. Would there be an easy way to fix it?

Thanks again for the package, and I would very much appreciate your help on this issue. Many thanks in advance!

[grantmcdermott]

Hmmm, that definitely doesn't seem right. Can you please paste your sessionInfo()results below.

Also, which version of Ubuntu are you on?

[romkhun]

Thanks for the quick response, Grant. The sessionInfo() is below. I am using Ubuntu 22.10.

[grantmcdermott]

Sorry @romkhun, I'm genuinely stumped for the moment. I can see that you have the latest version of the core packages (namely, etwfe and marginaleffects). Let me ask this: Did you install the CRAN version of etwfe or the development version?

[romkhun]

Actually, I tried both.

install.packages("etwfe", repos = "https://grantmcdermott.r-universe.dev") install.packages("etwfe")

Neither works for me.

[grantmcdermott]

Hi again @romkhun. So I just tested the package/example on an Ubuntu docker image and everything works as expected. Combined with the fact that all of my CI tests are passing (incl. two on Ubuntu), I think we can surmise that the problem has something to do with your particular system. Sorry, I know that's not helpful but I really can't reproduce this behaviour.

At this point, my next and possibly only best guess is that you have some kind of weird namespace clash. From your sessionInfo I can see that you have a bunch of other packages loaded. Can you please try in a fresh session with only etwfe loaded?

[grantmcdermott]

@romkhun See my previous comment. Please let me know if it works in a fresh session without a bunch of other libraries loaded at the same time. Otherwise, I'll have to close since I can't reproduce.

[grantmcdermott]

Closing since I cannot reproduce this behaviour.

[jsr-p]

I got the same bug using R version 4.2.3 and the newest versions of etwfe and marginaleffects. The issue seems to be with the insight package and the way it gets the modeldata from fixest; this is done inside the marginaleffects package. The returned data is passed through to get_contrast_data here and because the .Dtreat variable is not to be found inside modeldata the error is raised. Upgrading from insight version 0.19.1 to 0.19.7 solved it for me.

[grantmcdermott]

Upgrading from insight version 0.19.1 to 0.19.7 solved it for me.

Thanks @jsr-p, that's useful to know. I can easily add insight >= 0.19.7 to etwfe's Imports. But I'm wondering if this version requirement should rather be done upstream in marginaleffects, since this behavior may affect other similar cases. Any thoughts @vincentarelbundock?

[vincentarelbundock]

Sure, no problem: https://github.com/vincentarelbundock/marginaleffects/commit/42c49a119cf02b00af1076d29788bce064d5a0c0