I was using a function rq() from the quantreg package. Originally I fit a model with method="br" and got a warning message that the “solutions may be nonunique”. To suppress this warning, I tried fitting a model with different algorithm method="fn” and obtained the following results. However, the summary function on the model object (summary(fit)) yields different coefficients than fit$coefficients. Is there a way to make the summary function output a solution from the “fn” method rather than “br” method?

library(quantreg)

fit <- rq(y ~ x1 + x2 + x3, method="fn", data=dat, tau=0.5)
> fit$coefficients
(Intercept)          x1          x2          x3 
  9.4021951  25.1791775  14.1790548   0.2947553 

> summary(fit)

Call: rq(formula = y ~ x1 + x2 + x3, tau = 0.5, data = dat, method = "fn")

tau: [1] 0.5

Coefficients:
            coefficients lower bd upper bd
(Intercept)  9.35716      9.07227  9.85734
x1          25.17918     24.63725 26.15980
x2          14.22409     13.87955 14.43724
x3           0.24972     -0.28967  0.55808
Warning message:
In rq.fit.br(x, y, tau = tau, ci = TRUE, ...) : Solution may be nonunique

Sample data

set.seed(123)
n <- 100
x1 <- runif(100)
x2 <- sample(0:2, n, replace=TRUE)
x3 <- sample(c(0, 1), n, replace=TRUE)
y <- 10 + 25*x1 + 14*x2 + rnorm(n)
dat <- data.frame(y, x1, x2, x3)