Posterior_predict() for a specific value of predictor in rstanarm?


#1

I was wondering if it is possible to subset the posterior prediction for a specific predictor value (here hp = 93) from posterior_predict()?

fit <- stan_glm(mpg ~ hp, data = mtcars)
pred <- posterior_predict(fit)


#2
fit <- stan_glm(mpg ~ hp, data = mtcars)
pred <- posterior_predict(fit, newdata = data.frame(hp = 93))

#3

Thank you so much! Maybe this sounds like programming but is it possible to locate the exact location (i.e., which row which column) of the sample that is stored in pred below in the entire posterior_predict(fit)?

fit <- stan_glm(mpg ~ hp, data = mtcars)
pred <- posterior_predict(fit, newdata = data.frame(hp = 93))


#4

You mean

fit <- stan_glm(mpg ~ hp, data = mtcars)
pred <- posterior_predict(fit)[ , mtcars$hp == 93, drop = FALSE]

?


#5

I mean where is the sample stored in pred <- posterior_predict(fit, newdata = data.frame(hp = 93)) located in the larger sample from posterior_predict(fit)? Could we find the predictions for a specific value of predictor by knowing the row or a column number in the larger sample from posterior_predict(fit) instead of using newdata?


#6

Other than the ways outlined above, no.


#8

The columns are always in the same order as the rows of newdata (or, if not specified, the original data). I think that should be enough to find anything you want, right? Sorry if I’m misunderstanding.