Interpretation of the a zero-one inflated beta regression model estimated using brms

I have estimated the a zero-one-inflated beta regression model because my outcome variable has a lot of zeros and 1’s as well as some values between 0 and 1. The outcome variable is percent of milk sold to the market out of total amount produced. I would like to find out how I should interpret the output. I have regressors for the outcome variable (percentmilk), phi, zoi, and coi, terms. Any help will be appreciated. And thanks to @paul.buerkner for incorporating a zero-one-inflated beta regression into brms. Here is my simple code:





model1<-brm(f1, data=finale, chains = 2, cores=2, warmup = 2000, iter = 5000, thin = 5, control = list(adapt_delta = 0.99))

My issue is how to interpret the posterior means for independent variables under percentmilk, phi, zoi, and coi. Any help is welcome. Thanks.

I figured it out! No need for a response. Thank you!

Hello, nedson, i using the zoib model for the possession of driving license in R, i have tens variables, the model runs about 1H, but don’t finish. Isn’t it reasonable or not?

Hi Xiujuan, maybe you need to specify reasonable initial values. I experienced such a problem when you have many independent variables for all the ZOIB components. Try to restrict one or two of the model components as a function of a constant (e.g., phi~1 and/or coi~1) instead of including all the tens of your variables for all components. Then see what you get.

Hello! I also have an issue related to estimates interpretation (and, possibly calculation) in zero-one-inflated-beta regressions in @brms.
In model summary I continue to get reported links: mu = logit; phi = identity; zoi = identity; coi = identity, even though I defined them as logit, log, logit, logit, respectively.
For "mu"s, I have to apply inverse function to get interpretable results, but I am not sure how to deal with phi, zoi and koi. Is it possible that the inverse is applied before report generation or it is some kind of a bug? Can that occur due to some kind of model misspecification?