# [R] quantile regression with complex survey data

Donia Smaali Bouhlila sma.ali at fsjegj.rnu.tn
Tue Sep 16 23:39:28 CEST 2014

```Dear r Users,
I am new in r. I am trying to estimate regression quantiles in complex
surveys.I used these commands.

mydesign
<-svydesign(ids=~IDSCHOOL,strata=~IDSTRATE,data=TUNISIA,nest=TRUE,weights=~TOTWGT)
bootdesign <-  as.svrepdesign(mydesign,type="auto",replicates=150)
fit<withReplicates(bootdesign,quote(coef(rq(Math1~Female+Age+calculator+computer+desk
+dictionary+internet+work+Book2+Book3+Book4+Book5+Pedu1+Pedu2+Pedu3+Pedu4+Born1+Born2,tau=0.5,weights=.weights,
method="fn"))))

However, I don't get the results of the estimated coefficients.In
addition, I don't know how to calculate pseudo R squared.

When I type summary (fit) , I get the following results:

summary (fit)
Min. 1st Qu.  Median    Mean 3rd Qu.    Max.
-33.340  -6.358  12.200  38.800  17.510 645.700

For the pseudo R squared , I read the post [R] Pseudo R for Quant Reg
showing how to compute it:

rho <- function(u,tau=.5)u*(tau - (u < 0))
V <- sum(rho(f\$resid, f\$tau))

when I type

> rho <- function(u,tau=.5)u*(tau - (u < 0))
V <- sum(rho(fit\$resid, fit\$tau))

I get the following message
Error in fit\$resid : \$ operator is invalid for atomic vectors