2

I am learning mixed effects logistic regression from this link. In section "Analysis methods you might consider", the author listed several options:

  • Mixed effects logistic regression, the focus of this page.
  • Mixed effects probit regression is very similar to mixed effects logistic regression, but it uses the normal CDF instead of the logistic CDF. Both model binary outcomes and can include fixed and random effects.
  • Fixed effects logistic regression is limited in this case because it may ignore necessary random effects and/or non independence in the data.
  • Fixed effects probit regression is limited in this case because it may ignore necessary random effects and/or non independence in the data.
  • Logistic regression with clustered standard errors. These can adjust for non independence but does not allow for random effects.
  • Probit regression with clustered standard errors. These can adjust for non independence but does not allow for random effects.

I think I understand 1-4, but What is "Logistic regression with clustered standard errors"?

kjetil b halvorsen
  • 63,378
  • 26
  • 142
  • 467
Haitao Du
  • 32,885
  • 17
  • 118
  • 213
  • Check the answer [here](https://stats.stackexchange.com/questions/89810) for starters. – usεr11852 Apr 20 '17 at 17:33
  • @usεr11852 thanks for the link. Does it mean "run logistic regression anyway, but the "residual" will have patterns / clusters? " – Haitao Du Apr 20 '17 at 17:38
  • *Roughly speaking*: Yeah. (/me takes cover from angry econometricians). As a comment: Effectively all these SE estimators come from the realisation we might have heteroscedasticity. I would suggest you look first the Wikipedia lemma on [Heteroscedasticity-consistent standard errors](https://en.wikipedia.org/wiki/Heteroscedasticity-consistent_standard_errors) first. In R you might want to check examples using the packages [lmtest](https://cran.r-project.org/web/packages/lmtest/index.html) and [multiwayvcov](https://cran.r-project.org/web/packages/multiwayvcov/index.html). – usεr11852 Apr 20 '17 at 17:49
  • It was just a nudge :D Hopefully some user with expertise on the matter will come along and settle your question properly. (+1 Obviously) – usεr11852 Apr 20 '17 at 18:02
  • 1
    I don't think this has much to do w/ heteroscedasticity. Logistic regression essentially *always* has heteroscedasticty. Instead, the problem is that your observations are not independent. There can be adjustments to the SEs to address that. I should let others who know the topic better than I do answer, though. – gung - Reinstate Monica Apr 21 '17 at 16:04

0 Answers0