Question re brms: conditional_effects

I am fitting the model below with binary y (no/yes) and 3 categorical variables.
Then I use conditional_effects and I am not sure what I am getting. Am I supposed to get predicted posterior probabilities for yes per predictor level?

fit<- brm(formula =y~ x1 + x2+x3,  
                          family = bernoulli(link = "logit"),
                          warmup = 500,
                          iter = 50000,
                          chains = 4,
                          inits= "0",
                          seed = 123,

plot(conditional_effects(fit), ask = FALSE) 

Thank you,

Can you share a screenshot of what the plot(s) looks like?

Thank you Jim
The plot produced by the code
plot(conditional_effects(fit), ask = FALSE)
is below:
effect_plots.pdf (5.9 KB)
and it does not make much sense to me but it is very likely that I am missing something or doing the wrong thing

I used different coding and produced the desired plot-see example of 1 of the plots below :

predy <- base %>%
  data_grid(year,journorder,cont_aut) %>%
  add_fitted_draws(fit) %>%
  ggplot(aes(x = .value, y = year)) +
  stat_pointinterval(.width = c( .95),color="#E7B800",point_colour="black",
                     point_size=5,shape=16) +
  coord_flip() +
  xlab("Posterior Predicted Probability") +
  scale_x_continuous(breaks = seq(0, 1, 0.1))+
  scale_y_discrete(labels=c("0" = "2017", "1" = "2018","2" = "2019", 
                            "3" = "2020",
                            "4" = "2021"))+

the 3 desired plots combined
effects.pdf (6.6 KB)
thank you,

Something like this is what you would get, but from my understanding of conditional_effects() it is often going to plot the value of each predictor only at the reference class of the other predictors (so e.g., for that journorder plot among the 3 conditional effects plots, it is probably actually plotting the different levels of journal order at whatever is the reference class for the other two variables. I’m not 100% sure about that though).

If you want to see predictions I would suggest checking out the function posterior_epred(). You could specify as the newdata argument a tibble that contains all combinations of the different predictors (e.g., by using expand_grid(“journorder” = c(1, 2, 3, 4, 5), “cont_auth” = bla bla…) ). The output of posterior_epred will be a matrix that has the posterior distribution for the predicted values of each variable at each level of the other variables.

You might also be interested to see if the package emmeans works with binomial outcomes in brms, as it is a handy way of generating marginal effects estimates.

Apologies if this is no help, and feel free to ignore if that last bit of code you did already solved your issue!

Thank you Jim
I think I have solved my problem but I will check also your suggestion