[R] trouble getting table of coeffs with quantreg with fixed effects

[Steve McDonald]

I see. Thank you, David. One more question: How can one obtain bootstrapped standard errors for these regression coefficients. I tried the boot.rq procedure, but the results do not look right. 

Steve

--
Steve McDonald
Assistant Professor of Sociology
North Carolina State University
Phone: 919-515-9028
Office: 343 in 1911 bldg
http://www4.ncsu.edu/~sjmcdona/


>>> David Winsemius <dwinsemius at comcast.net> 7/20/2010 8:18 AM >>>

On Jul 19, 2010, at 9:37 PM, Steve McDonald wrote:

> I'm a new user, so my apologies for what is likely a dumb question...
>
> I am having a hard time getting a table of regression results when  
> using Koenker's code for quantile regression with fixed effects (http://www.econ.uiuc.edu/~roger/research/panel/rq.fit.panel.R 
> ). I use the example data parameters that Koenker provides (see  
> below).
>
> m <- 3
> n <- 10
> s <- rep(1:n,rep(m,n))
> x <- exp(rnorm(n*m))
> X <- cbind(1,x)
> u <- x*rnorm(m*n) + (1-x)*rf(m*n,3,3)
> a <- rep(rnorm(n),rep(m,n))
> y <- a + u
> fit <- rq.fit.panel(X,y,s)
>
> When I use "summary(fit)", I get the following output...
>     Length Class  Mode
> coef 16     -none- numeric
> ierr  1     -none- numeric
> it    1     -none- numeric
> time  1     -none- numeric
>
> When I use "list(fit)", I can't tell which coefficients correspond  
> to which parameters...
> [[1]]$coef
> [1]  7.701321 -2.608782  7.664906 -1.866059  7.894542
> [6] -1.878973 -6.117778 -1.570840 -5.380230 -5.815231
> [11] -7.191127 -5.671467 -1.712052 -5.055224 -5.623140
> [16] -5.617846
>
> Is there a way to get a table of results or a way of discerning what  
> these coeffs refer to? Thanks in advance.

Did you read the documentation in the rq.fit.panel function?

" 2.  On return the coefficient vector has m*p + n elements where m is  
the number
+ #	quantiles being estimated, p is the number of colums of X, and n  
is the
+ #	number of distinct values of s.  The first m*p coefficients are the
+ #	slope estimates, and the last n are the "fixed effects""
>
-- 

David Winsemius, MD
West Hartford, CT