Questions on the calculation of effective degrees of freedom in weakivtest

[Karl Weierstrass]

I am running regressions using

ivreg2

, and testing weakness of my IV using

weakivtest

, which implements the well-known heteroskedasticity-robust test proposed by Montiel Olea and Pflueger (2013). I noticed that the critical values given by weakivtest has always remained constant, which seemed strange since I have changed the sample size for several times. Also, the critical values seemed not to be dependent on my choice to partial the constant out.

My regression is a simple IV-LS, with only one dependent variable, one independent variable and one instrument. It looks like this:

ivreg2 y (x = z), first noconst robust
weakivtest

This phenomenon seems odd to me because in the ordinary F-test the critical values depends on the degrees of freedom, which in turn depends on the sample size and the inclusion of an intercept. I know little econometrics, so it is hard for me to quickly go through Montiel Olea et al. (2013), especially the part where they calculate the effective degrees of freedom using an algorithm too complicated for me to understand. I wonder whether it is normal that the critical values (and effective degrees of freedom) remained the same when I changed the sample size or excluded the intercept? Or maybe

weakivtest

uses a default degree of freedom of 1? I am not sure whether this could be related to my problem. Can anyone please explain to me as plain as possible? Thanks in advance!

Also, I have read a 2020 post on this forum mentioning that

weakivtest

does not support IV regression without constant, but it seems that the authors has updated the code to allow for noconstant.

[Karl Weierstrass]

I am new to this forum and have not figured out how to insert inline codes. I reformat my question here so that it is more readable.

I am running regressions using ivreg2, and testing weakness of my IV using weakivtest, which implements the well-known heteroskedasticity-robust test proposed by Montiel Olea and Pflueger (2013). I noticed that the critical values given by weakivtest has always remained constant, which seemed strange since I have changed the sample size for several times. Also, the critical values seemed not to be dependent on my choice to partial the constant out.

My regression is a simple IV-LS, with only one dependent variable, one independent variable and one instrument. It looks like this:

ivreg2 y (x = z), first noconst robust
weakivtest

This phenomenon seems odd to me because in the ordinary F-test the critical values depends on the degrees of freedom, which in turn depends on the sample size and the inclusion of an intercept. I know little econometrics, so it is hard for me to quickly go through Montiel Olea et al. (2013), especially the part where they calculate the effective degrees of freedom using an algorithm too complicated for me to understand. I wonder whether it is normal that the critical values (and effective degrees of freedom) remained the same when I changed the sample size or excluded the intercept? Or maybe weakivtest uses a default degree of freedom of 1? Can anyone please explain to me as plain as possible? Thanks in advance!

Also, I have read a 2020 post on this forum mentioning that weakivtest does not support IV regression without constant, but it seems that the authors has updated the code to allow for noconstant. I am not sure whether this could be related to my problem.