Marginal_effects data from brm_multiple (newbie question)

I’ve successfully used brm_multiple for ordinal regression of data from a national survey, and the marginal_effects plot is great. Apologies if this should be obvious, but since I haven’t found an answer, how can I see (and report) the actual data, the marginal effects represented by the points on the plot?

I created the plot with
based on the example in the Bürkner/Vuorre tutorial.
“screens_model_1” is the brmsfit_multiple object.
“year” is an indicator variable in the data.


Perhaps this will get you what you want:

ce <- conditional_effects(screens_model_1, "year", categorical = TRUE)

plot(ce, points = TRUE)

Thanks (and sorry for the delay–my RStudio spent a full 24 hours processing a model that didn’t work in the end, so I couldn’t check out your suggestion.)
This gave me a plot similar to the one I have. What I’m trying to find is numerical data for each of the points so I can report it, rather than writing “look at the pink point about a quarter of the way below 0.1 on the Y axis , and just to the right of 2007 on the X axis.”

You mean you want summaries, such as posterior means and 95% intervals? Use fitted().

Thanks for hanging in there with me! No, I have the summary data, but the parameters don’t provide the information I need. I am looking at a plot (tried to upload it, hope that worked!) with years (ordinal) on the X axis and probability (0 to 0.275) on the Y axis. Each point on the plot represents the marginal probability of a student choosing one of 7 answer choices to a multiple-choice item. Intuitively, I can see that the marginal probability of each answer choice, (as opposed to the overall year-to-year change indicated in the summary) increases or declines from one year to the next. What I want is the underlying data: What was the marginal probability of answer choice C in 2011 vs 2013?screens_model1 marginal effects
The numbers must be somewhere, else how were they plotted?

It’s not entirely clear what you mean by “data,” in this context. In the terms I would use, the only “data” are the ones you already have at hand–they’re what you fed into the model. If you’re looking for the MCMC posterior draws, perhaps the posterior_samples() function will give you what you need.


@rmcuny, does:


give you what you’re looking for?