I'm trying to fit a GLMM using the package glmmTMB in R. The model has a negative binomial distribution. The fixed effect "Treatment" has two levels (low or high), "Sampling" is the number of consecutive times a colony has been sampled (14 levels). "Size" is a covariate that account for the size of the individuals, and the colony and individual ID are included as random effects. I have 6 replicates per treatment.

Model <- glmmTMB(cell_con ~ treatment * sampling + size + (1|colony/ind_ID),
                 family = "nbinom2", data=countI)

The model fits when I only include the main effects, but when I include interactions between terms I get the following warning. I'm actually mainly interested in the interaction between treatment and sampling.

"Warning message: In fitTMB(TMBStruc) : Model convergence problem; non-positive-definite Hessian matrix. See vignette('troubleshooting')"

I already check the Troubleshooting page for glmmTMB, but unfortunately I'm not that well versed in statistics to solve the problem with that information


PD: this is the first time I post a questions here. Let me know if I should post more information about the experiment. I can also send the complete code and data.

  • 1
    $\begingroup$ Your model is having problems converging, 1) are your variables highly correlated? 2) Do you have enough observations to estimate all the terms? 3) Try a different optimizer. $\endgroup$ Commented Dec 20, 2018 at 6:33
  • 1
    $\begingroup$ Perhaps you don’t need the nested random effects. $\endgroup$ Commented Dec 21, 2018 at 7:49
  • $\begingroup$ see troubleshooting with glmmTMB - cran.r-project.org/web/packages/glmmTMB/vignettes/… $\endgroup$ Commented Dec 10, 2019 at 4:43
  • $\begingroup$ @user2974951 What were you referring to when you said to try a different optimizer? I don't see an optimizer specified in the original model. $\endgroup$ Commented Sep 16, 2020 at 19:08
  • $\begingroup$ @SusieDerkins Admittedly I don't know this specific package so there may well not be such an option, but I know that other more standard packages do have this option. So maybe the answer is try a different package. $\endgroup$ Commented Sep 17, 2020 at 5:59

1 Answer 1


the following site reports the troubleshooting you are mentioning: https://cran.r-project.org/web/packages/glmmTMB/vignettes/troubleshooting.html

the following point can be the problem when you are getting the message.These problems are most likely:

  • when a model is overparameterized (i.e. the data does not contain enough information to estimate the parameters reliably)
  • when a random-effect variance is estimated to be zero, or random-effect terms are estimated to be perfectly correlated ("singular fit": often caused by having too few levels of the random-effect grouping variable)
  • when zero-inflation is estimated to be near zero (a strongly negative zero-inflation parameter)
  • When dispersion is estimated to be near zero
  • When complete separation occurs in a binomial model: some categories in the model contain proportions that are either all 0 or all 1

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.