about demeaned covariates

[pedrohcgs]

Hi @grantmcdermott

Thanks for all the work here. I am playing with the package, and I want to know if the std errors calculated in etwfe consider that the covariates' sample mean is random and estimated.

I mean, you are recentering all the covariates before you interact them, but are you treating the sample mean as the population mean? In the sampling approach adopted in Wooldridge, you would need to account for this extra randomness. But that would not be the case in the finite population setup a la Lin (2013).

Thanks again,
Pedro

[grantmcdermott]

Sorry for the slow reply @pedrohcgs . You caught me while I was away....

Oof. Honestly, it's been a while since I looked at this code, but the short answer is 'no'. I didn't do any special adjustments for sampling error. (I seem to recall Jeff Wooldridge asking me about this back when I first released the package.) I'm trying to remember now, but I don't think jwdid does this adjustment either and I was initially benchmarking my results against this latter package. OTOH, I also did some benchmarking against Wooldridge's manual implementation and IIRC my standard errors and his were typically equivalent until 4 significant digits. So I'm not sure that it will make much difference in practice, if we're already accounting for clustering etc. (Could be wrong!)

I'm currently juggling few other OSS projects (outside of "office hours" unfortunately). But I'll aim to take a deeper look at this when I can. Appreciate any additional thoughts you might have, though ;-)

[grantmcdermott]

Quick update: I believe that this requires an adjustment (or allowance for manual adjustments) on the marginaleffects side. Tracking upstream: vincentarelbundock/marginaleffects#1240

[pedrohcgs]

I have a feeling that if you do not recenter, marginal effects would take care of it. But I am not 100% sure.

…

---------------------------------------------------- *Pedro H. C. Sant'Anna* *https://psantanna.com <https://psantanna.com>*

---------------------------------------------------- Warning: This email may contain confidential or privileged information intended only for the use of the individual or entity to whom it is addressed. If you are not the intended recipient, please understand that any disclosure, copying, distribution, or use of the contents of this email is strictly prohibited. ----------------------------------------------------

On Mon, Oct 21, 2024 at 16:53 Grant McDermott ***@***.***> wrote: Quick update: I believe that this requires an adjustment (or allowance for manual adjustments) on the marginaleffects side. Tracking upstream: vincentarelbundock/marginaleffects#1240 <vincentarelbundock/marginaleffects#1240> — Reply to this email directly, view it on GitHub <#48 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABE7346N5SG6WJ3J6VZ7THTZ4VZVZAVCNFSM6AAAAABQELOT7CVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDIMRXHAYTCNJQG4> . You are receiving this because you were mentioned.Message ID: ***@***.***>