Abadie et al (2023) standard errors?

[kendonB]

Hi, do you have any plans to implement the standard error estimators from https://academic.oup.com/qje/article/138/1/1/6750017?rss=1 ?

I haven't yet derived the continuous, multivariate version of the CCV estimator. I have emailed Prof. Wooldridge to see if they have it written out somewhere.

[kendonB]

I also contacted Achim Zeileis who correctly pointed out that it can't be computed using a traditional sandwich approach so would need to live somewhere like fixest.

[waynelapierre]

sounds like a great feature to have!

[kendonB]

The answer from JW was that they don't yet have the general case derived, he believes it will generalise, and he believes that others are working on it.

[s3alfisc]

I'd be happy to work on this eventually @kendonB - maybe we could start with an external package which allows to compute the Abadie et al standard errors as a post-estimation command, and then prepare a PR for fixest?

[s3alfisc]

I think I'll give this a go and try to integrate the AAIW inference procedures as post estimation commands into fwildclusterboot over the next weeks (PR here), unless you'd want to implement them in 'native fixest' in the nearer future, @lrberge ?

[kendonB]

Hi @s3alfisc, please go for it - the general case for the CCV estimator doesn't look trivial to derive by analogy and JW has requested that people contact him if they try. The TSCB seems like it would be easier to generalise to the multivariate continuous case. Happy to test and, if I can, review.

[fa-rzam]

Hi @s3alfisc, I wonder if you might have implemented the CCV or TSCB already? Thanks!

[s3alfisc]

Hi, not yet, I started a while back but never then always other priorities popped up! If I find some free time this weekend, I'll try to see how far I can get with this =)