Thresholds for ordinal brms model in emmeans

Dear community,

I wonder why an ordinal mixed effect model in brms does not have thresholds in emmeans/emmip? In my case, the outcome is some mean probability over all thresholds, i.e. response range. Is it possible to get the marginal probability for each threshold?

For example, in the package ordinal all thresholds are coded as “cut” variables and can be identified in ref_grid in emmeans example. How can I identify thresholds in emmeans providing my model is fitted in brms? Maybe I miss some details in coding?

Note: I do not use conditional_effects here, as I would like to estimate marginal mean at different offset values and emmip is more convenient for this.
The fitted model is as following:

f = brmsformula(count1~ ba_sqrt + dbh_log + species.p + species.m + status.p +
position.m + y + offset(log(count_dead)) + (1+y|subplot_id), center=TRUE) +
lf(disc ~ 0 + y + status.p + species.p, cmc = FALSE)

fit=brm(f, data=d1, family=cratio(link="probit"), prior=p, 
save_pars=save_pars(all = TRUE), cores  = 2, iter = 800 + 800, 
warmup = 800, chains = 1, seed=123,
sample_prior="yes", silent=TRUE, open_progress=FALSE, 
control = list(adapt_delta = 0.85, max_treedepth = 10))

# getting the grid in emmeans:
con = ref_grid(fit, at = list(status.p = "alive", ba_sqrt=1.41, 
dbh_log=seq(1.6, 3, by=0.1), bias.adj = TRUE))

# plotting
emmip(con, species.p ~ dbh_log | position.m, type="response", 
offset = log(4)) + theme(legend.position = "top") + 
geom_ribbon(data=b, aes(x=dbh_log, y=yvar, ymin=LCL, ymax=UCL, fill=species.p), alpha=0.25, inherit.aes = FALSE) + 
geom_line(data=b, aes(x=dbh_log, y=yvar, color=species.p), size=1, inherit.aes = FALSE)+ labs(x="log(dbh)

# outcome:
             species.p dbh_log    position.m      yvar       LCL       UCL
1      Acer platanoides     1.6 adjacent_dead 0.6778162 0.5495729 0.7993685
2   Acer pseudoplatanus     1.6 adjacent_dead 0.7822773 0.6304033 0.8995334
3       Fagus sylvatica     1.6 adjacent_dead 0.8808129 0.7851305 0.9499130
4          Ulmus glabra     1.6 adjacent_dead 0.5135319 0.3046328 0.6891265


I would appreciate any idea or hints very much.

  • Operating System: Windows 10
  • brms Version: 2.14.0