Extract coefficient with p from a quantile regression by group with dplyr

Question

I would like to extract p-values and coefficients from a series of quantile regressions made with a grouping variable. I mostly use dplyr to manipulate data frames so a would like dplyr solution.

require(quantreg)
data("engel")
require(dplyr)
engel$grp <- trunc(runif(nrow(engel), min=0, max=3))

group_by(engel,grp) %>% do(summary(rq(foodexp~income,data=.,tau=c(.05, .25, .5, .75, .95)),se="boot"))

This result in a error

Error: Results are not data frames at positions: 1, 2, 3

I tried another version doing first the models and later the summary

rqm <- group_by(engel,grp) %>% do(mdl=rq(foodexp~income,data=.,tau=c(.05, .25, .5, .75, .95)))
summarise(rqm, coef(summary(mdl,se="boot")))

which also result in error

Error: not a vector

Answer 1

It's a mess, but it works:

library(dplyr)
group_by(engel,grp) %>% 
      do(as.data.frame(do.call(rbind,
          lapply(summary(rq(foodexp~income,data=.,tau=c(.05, .25, .5, .75, .95)), se="boot"), coef)
            ), row.names = NULL))

You would then want to label the rows with the tau values and if it is a coef or p value. I'll leave that to you.