Open strengejacke opened 5 years ago
A less complex example, taken from ?brms::mo
:
income_options <- c("below_20", "20_to_40", "40_to_100", "greater_100")
income <- factor(sample(income_options, 100, TRUE),
levels = income_options, ordered = TRUE)
mean_ls <- c(30, 60, 70, 75)
ls <- mean_ls[income] + rnorm(100, sd = 7)
dat <- data.frame(income, ls)
# fit a simple monotonic model
fit1 <- brm(ls ~ mo(income), data = dat, iter = 500, chains = 1)
broom.mixed::tidy(fit1)
See also https://github.com/paul-buerkner/brms/issues/543 for a brms issue I just opened.
And simo_mo
is the prefix for the simplex parameters, right?
simo_ is the prefix for simplexes of monotonic effects
Am So., 4. Nov. 2018, 14:41 hat Daniel notifications@github.com geschrieben:
And simo_mo is the prefix for the simplex parameters, right?
— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/bbolker/broom.mixed/issues/45#issuecomment-435670689, or mute the thread https://github.com/notifications/unsubscribe-auth/AMVtAD3KUQdhOSpBac6QQKwee9QTVdwSks5uru6igaJpZM4YNQTL .
This is a bit offtopic, but as far as I understand, for non-gaussian models, I would not transform (e.g. exponentiate) the simplex parameters, as these indicate proportions of the range, which is indicated by the coefficient of the simplex term. But what would you recommend regarding this coefficient?
I've implemented support for monotonic effects in sjPlot::tab_model()
, see example for the two models I posted above:
tab_model(m, fit3)
Left is the poisson-model, right the linear model. "income" is exponentiated in the left model.
I think we should discuss this somewhere else. Perhaps on https://discourse.mc-stan.org/?
Yes, will do. I thought it might be semi related because some tidy methods allow to exponentiate the estimates, but it'll be too much off topic.
Paul has published a paper, which describes how to model monotonic effects in brms. These terms (indicated in the formula with
mo()
) are not shown in the output oftidy()
(using broom.mixed 0.2.3). The data resp. the model does not make much sense with theincome
variable, this is just for demonstration purposes.Created on 2018-11-04 by the reprex package (v0.2.1)