0

I have fitted the following model:

cn_sign.mod3 <- glmer(sign ~ time + gmc_secident + time*gmc_cn + 
    (time|id), US_mlm, family = "binomial", 
     control = glmerControl(check.nobs.vs.nRE = "ignore"))

The model itself works but I have problems calculating the odds ratio and confidence interval. I have tried using lme4::confint.merMod(cn_sign.mod3) and tried boosting devtol = but even with devtol = 1e-01 (I went down stepwise) it doesn't work (I took this from here: Error when estimating CI for GLMM using confint())

For the odds ratio I have tried:

cc_cn_sign.mod3 <- confint(cn_sign.mod3, parm = "beta_")
ctab_cn_sign.mod3 <- cbind(est=fixef(cn_sign.mod3),cc_cn_sign.mod3)
rtab_ccn_sign.mod3 <- exp(ctab_cn_sign.mod3)
print(rtab_cn_sign.mod3, digits = 3)

as well as

or.glmer(US_mlm, cn_sign.mod3, complex)

(from https://rdrr.io/github/mjwestgate/sppairs/man/or.glmer.html)

but nothing seems to work. My DV is very uneven (with 1 = N of 15, and 0 = N of 89) and I am afraid that this influences the functionality of this model?

Ben Bolker
  • 211,554
  • 25
  • 370
  • 453
Carina Hörst
  • 1
  • 1
  • can you edit your question to give the specific error encountered? (the linked answer suggests what it is, but it would be nice if your question were self-contained) – Ben Bolker Jan 22 '21 at 22:26
  • Is there any chance we can have a [mre]? In this case it probably means providing your full data set ... – Ben Bolker Jan 22 '21 at 22:28

0 Answers0