Using Robust Standard Errors lme4

Hi Suresh,
I just tried to find an example with glmer(), but it indeed looks like the
clubSandwich package only supports lmer() models, not glmer(). So my
suggested approach unfortunately doesn't seem to work.

Best
Daniel

Am Di., 9. Nov. 2021 um 17:24 Uhr schrieb Suresh N Neupane <
sneupane2 at student.gsu.edu>:

Hi Daniel,
Thank you so much for your suggestions. I tried that but it gives me an
error:

Error in UseMethod("estfun") :
  no applicable method for 'estfun' applied to an object of class
"c('glmerMod', 'merMod')"

My DV is death (0/1) and IV other variables:
Model: (with random county effects):
fit <- glmer(death ~ (1|county_fips_code) +white + black + pop.density ,
family = binomial("logit"), nAGQ=0, data = data.thesis)
 I tried the code you sent and used "county_fips_code" as my cluster
variable but it says did not find "county_fips_code".
I was able to use parameters::model_parameters(fit) to extract some info
(This is great!) such as;
 Parameter     |  Log-Odds |   SE |  95% CI |   z |     p

Thank you so much again,
I'd really appreciate if you would have more input on this.
Sincerely,
Suresh




------------------------------
*From:* Daniel L?decke <d.luedecke at uke.de>
*Sent:* Tuesday, November 9, 2021 4:32 AM
*To:* Suresh N Neupane <sneupane2 at student.gsu.edu>
*Cc:* r-sig-mixed-models at r-project.org <r-sig-mixed-models at r-project.org>
*Subject:* AW: [R-sig-ME] Using Robust Standard Errors lme4

Hi Suresh,

you could try the "parameters" package, which should (easily) return
different types of robust standard errors, including cluster-robust
standard
errors for mixed models:


https://nam12.safelinks.protection.outlook.com/?url=https%3A%2F%2Feasystats.github.io%2Fparameters%2Farticles%2Fmodel_parameters_robust.html&amp;data=04%7C01%7Csneupane2%40student.gsu.edu%7Ca99f7d4a171f419a724a08d9a363ee14%7C704d822c358a47849a1649e20b75f941%7C0%7C0%7C637720471875578865%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C1000&amp;sdata=DOoXhh%2FzuFnNc6ZTAb9hry9%2FSB7QD%2F7bEYX%2BaHLQBhE%3D&amp;reserved=0

Maybe you try this out on a very small subset of your model to see if it
works. Also, since the "model_parameters()" function includes random
effects
variances by default, you may set the effects-argument to "fixed" to speed
up computation. The call to the function could then look like something
like
this:

model_parameters(
  model,
  effects = "fixed",
  robust = TRUE,
  vcov_estimation = "CL",
  vcov_type = "HC1",
  vcov_args = list(cluster = <your cluster variable from the data>)
)


Best
Daniel

-----Urspr?ngliche Nachricht-----
Von: R-sig-mixed-models <r-sig-mixed-models-bounces at r-project.org> Im
Auftrag von Suresh N Neupane via R-sig-mixed-models
Gesendet: Montag, 8. November 2021 12:49
An: r-sig-mixed-models at r-project.org
Betreff: [R-sig-ME] Using Robust Standard Errors lme4

Dear all,
I am using lme4 package to run a hierarchical logistic regression model
(with random county effect) for my binomial dependent variable, COVID death
(0/1). This is a very large dataset with ~ 8 million observations.
I need to find the robust standard error but was not sure about any
package.
I tried robustlmm (although my DV is not continuous) and got this error
when
used MerDeriv (sandwich):
Error in UseMethod("estfun") :
no applicable method for 'estfun' applied to an object of class
"c('glmerMod', 'merMod')"
The real tricky part is it takes several hours to compute because of the
dataset length.
All I need is to get robust standard error values.
I already fit the models:
fit <- glmer(death ~ (1|county_fips_code) + IV).
Thank you so much,
Suresh Neupane

--