# How to get p-values for the zero-inflated part of a model?

I am having a problem that I am not finding an answer anywhere online.

I used glmmTMB to fit a mixed-model. My data is zero-inflated and I am including 3 variables in the zero-inflated formula.

I was using the Anova in the car package to extract p-values for each of my variables (I know that there are some limitations to it!). However, this only allows me to extract the p-values for the conditional model, and not for the zero-inflated or hurdle model. The summary of the model gives me some p-values but I have categorical variables and, therefore, I get a value for each level of the factors.

Is there any way I can get p-values for the zero-inflated part of the model? And if not, what would be the best way to report a zero-inflated or a hurdle model?

Any help would be greatly appreciated.

If you check the vignette, you can specify the component using component = :

library(glmmTMB)
library(car)

m1 <- glmmTMB(count ~ spp + mined + (1|site),zi=~mined,family=poisson, data=Salamanders)

car::Anova(m1,component="cond")
Analysis of Deviance Table (Type II Wald chisquare tests)

Response: count
Chisq Df Pr(>Chisq)
spp   125.625  6  < 2.2e-16 ***
mined  22.238  1  2.408e-06 ***
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

car::Anova(m1,component="zi")
Analysis of Deviance Table (Type II Wald chisquare tests)

Response: count
Chisq Df Pr(>Chisq)
mined 34.509  1  4.243e-09 ***
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1


It is written:

car::Anova constructs type-II and type-III Anova tables for the fixed effect parameters of the conditional model (this might work with the fixed effects of the zero-inflation or dispersion models, but has not been tested)

Don't know if it is tested now..

• Thank you @StupidWolf. Not sure what to do with the "not tested yet" part, though... Jun 29, 2020 at 9:43
• You can simulate your model under the null and see whether anova test makes sense Jun 29, 2020 at 10:31
• unfortunately I do not quite understand what you mean or how to do it. Would it be possible to be more specific? For instance, using your example above... Jun 29, 2020 at 12:07