Vglm in Rstan

Modeling
1

any idea how i can reproduce multinomial logit model such as vglm() in Rstan or in rstanarm(no random effects).From my web search i only found ordinal regression.

2

It isn’t in rstanarm. Writing a Stan program to do it is not that hard once you pick the form of the likelihood you want to use (and priors).

3

For me is very difficult.I am not a computer scientist.Thanks a lot for your reply.I appreciate that.

4

You can do multinomial regression with brms. There’s an example thread here; searching the boards for “brms multinomial” will turn up more examples.

5

thanks but when i run the code in the example page in R i take an error:Error: Argument ‘family’ is invalid. for multinomial()

6

syntax error is family = “multinomial”.Thanks

7

It would help to see

  1. the code you actually ran, and
  2. the exact error message.

You should copy-and-paste these in your posts, rather than just presenting an ambiguous or possibly misleading summary.

8 
race   = gl(2,2,labels   = c("white","black"));race
gender = gl(2,1,labels = c("female","male"));gender
y1     = c(371,250,64,25)
y2     = c(49,45,9,5)
y3     = c(74,71,15,13)
afterlife = data.frame(race,gender,y1,y2,y3)
afterlife
afterlife$size = with(afterlife, y1 + y2 + y3)
afterlife$y = with(afterlife, cbind(y1, y2, y3))
prior <- prior(normal(0, 10), "b", dpar = muy2) +
  prior(cauchy(0, 1), "Intercept") +
  prior(normal(0, 2), "Intercept", dpar = muy3)
fit <- brm(bf(y | trials(size)  ~ 1, muy2 ~ race+gender), data = afterlife, 
           family = "multinomial", prior = prior)
print(fit)
plot(fit)
predict(fit)
9

But the output that i get its completely different.In brm

 y | trials(size)  ~ 1

what does it mean?that the first category is the reference?

race   = gl(2,2,labels   = c("white","black"));race
gender = gl(2,1,labels = c("female","male"));gender
y1     = c(371,250,64,25)
y2     = c(49,45,9,5)
y3     = c(74,71,15,13)
afterlife = data.frame(race,gender,y1,y2,y3)
fit2 <- vglm(cbind(y1,y2,y3) ~ gender + race, family=multinomial,
            data=afterlife)
summary(fit2)

what the right modelling for these data?