# Question of "Quantile Regression for Longitudinal Data"

9 messages
Open this post in threaded view
|

## Question of "Quantile Regression for Longitudinal Data"

 Hi, I am trying to estimate a quantile regression using panel data. I am trying to use the model that is described in Dr. Koenker's article. So I use the code the that is posted in the following link: http://www.econ.uiuc.edu/~roger/research/panel/rq.fit.panel.RHow to estimate the panel data quantile regression if the regression contains no constant term? I tried to change the code of rq.fit.panel by delect "X=cbind(1,x)" and would like to know is that correct ? Thanks I really would appreciate some suggestions. Best Helen Chen
Open this post in threaded view
|

## Re: Question of "Quantile Regression for Longitudinal Data"

 This is a nontrivial problem. This comes up often on the Statalist (-qreg- is for  cross-section quantile regression): " You want to fit a plane through the origin using the L-1 norm. This is not as easy as with L-2 norm (LS), as it is more than a matter of dropping a constant predictor yet otherwise using the same criterion of fit. You are placing another constraint on a problem that already does not have a closed-form solution, and it does not surprise me that -qreg- does not support this. " (N.J. Cox) http://www.stata.com/statalist/archive/2007-10/msg00809.htmlYou will probably have to program this by hand. Note also the degeneracy conditions in Koenker (2003, pg. 36--). I am not sure how this extends to panel data though. References: @book{koenker2005qre,   title={{Quantile Regression; Econometric Society Monographs}},   author={Koenker, R.},   year={2005},   publisher={Cambridge University Press} } T On Sun, Apr 26, 2009 at 8:24 AM, Helen Chen <[hidden email]> wrote: > > Hi, > > I am trying to estimate a quantile regression using panel data. I am trying > to use the model that is described in Dr. Koenker's article. So I use the > code the that is posted in the following link: > > http://www.econ.uiuc.edu/~roger/research/panel/rq.fit.panel.R> > How to estimate the panel data quantile regression if the regression > contains no constant term? I tried to change the code of rq.fit.panel by > delect "X=cbind(1,x)" and would like to know is that correct ? > > > Thanks > I really would appreciate some suggestions. > Best > Helen Chen > -- > View this message in context: http://www.nabble.com/Question-of-%22Quantile-Regression-for-Longitudinal-Data%22-tp23239896p23239896.html> Sent from the R help mailing list archive at Nabble.com. > > ______________________________________________ > [hidden email] mailing list > https://stat.ethz.ch/mailman/listinfo/r-help> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html> and provide commented, minimal, self-contained, reproducible code. > -- To every ω-consistent recursive class κ of formulae there correspond recursive class signs r, such that neither v Gen r nor Neg(v Gen r) belongs to Flg(κ) (where v is the free variable of r). ______________________________________________ [hidden email] mailing list https://stat.ethz.ch/mailman/listinfo/r-helpPLEASE do read the posting guide http://www.R-project.org/posting-guide.htmland provide commented, minimal, self-contained, reproducible code.
Open this post in threaded view
|

## Re: Question of "Quantile Regression for Longitudinal Data"

 I was trying to resist responding to this question since the original   questioner had already been admonished twice  last october about asking questions on R-help about posted code that was not only not a part of R-base, but not even a part of an R package.  But the quoted comment about Stata is too enticing a provocation to resist. First, it should be said that omitting intercepts in any regression   setting should be undertaken "at one's peril"  it is generally a very dangerous activity, somewhat akin to fitting interactions without main effects,   but if there is a good rational for it, it is no different in principle for   median regression than for mean regression.  It may well be that Stata   prohibits this sort of thing out of some sort of paternalistic motive, but in R   the usual formula convention  y ~ x1 + x2 -1 suffices.  Of course it   situations in which such a formula is used for several quantiles it should be   understood that it is forcing each conditional quantile function through the origin effectively implies that the conditional distribution degenerates to a   point mass  at the origin. Second,  I would like to remark that "closed-form solutions" are in   the eye of the beholder, and many people who can recall the infamous formula:         betahat = (X'X)^{-1} X'y would be hard pressed to  dredge up enough linear algebra to use the formula for anything more than the bivariate case on  the proverbial desert island  without the aid of their trusty  laptop "Friday". Finally,  cbind(1,x) does introduce an intercept in the code   originally asked about, so if you don't want an intercept don't do that, but be sure   that that is really want you want to do. url:    www.econ.uiuc.edu/~roger                Roger Koenker email   [hidden email]                       Department of Economics vox:    217-333-4558                            University of Illinois fax:    217-244-6678                            Champaign, IL 61820 On Apr 26, 2009, at 6:35 AM, Tirthankar Chakravarty wrote: > This is a nontrivial problem. This comes up often on the Statalist > (-qreg- is for  cross-section quantile regression): > " > You want to fit a plane through the origin using the L-1 norm. > This is not as easy as with L-2 norm (LS), as it is more > than a matter of dropping a constant predictor yet otherwise using the > same criterion of fit. You are placing another constraint on a > problem that already does not have a closed-form solution, > and it does not surprise me that -qreg- does not support this. > " (N.J. Cox) > http://www.stata.com/statalist/archive/2007-10/msg00809.html> > You will probably have to program this by hand. Note also the > degeneracy conditions in Koenker (2003, pg. 36--). I am not sure how > this extends to panel data though. > > References: > @book{koenker2005qre, >  title={{Quantile Regression; Econometric Society Monographs}}, >  author={Koenker, R.}, >  year={2005}, >  publisher={Cambridge University Press} > } > > T > > On Sun, Apr 26, 2009 at 8:24 AM, Helen Chen <[hidden email]>   > wrote: >> >> Hi, >> >> I am trying to estimate a quantile regression using panel data. I   >> am trying >> to use the model that is described in Dr. Koenker's article. So I   >> use the >> code the that is posted in the following link: >> >> http://www.econ.uiuc.edu/~roger/research/panel/rq.fit.panel.R>> >> How to estimate the panel data quantile regression if the regression >> contains no constant term? I tried to change the code of   >> rq.fit.panel by >> delect "X=cbind(1,x)" and would like to know is that correct ? >> >> >> Thanks >> I really would appreciate some suggestions. >> Best >> Helen Chen >> -- >> View this message in context: http://www.nabble.com/Question-of-%22Quantile-Regression-for-Longitudinal-Data%22-tp23239896p23239896.html>> Sent from the R help mailing list archive at Nabble.com. >> >> ______________________________________________ >> [hidden email] mailing list >> https://stat.ethz.ch/mailman/listinfo/r-help>> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html>> and provide commented, minimal, self-contained, reproducible code. >> > > > > -- > To every ω-consistent recursive class κ of formulae there correspond > recursive class signs r, such that neither v Gen r nor Neg(v Gen r) > belongs to Flg(κ) (where v is the free variable of r). > > ______________________________________________ > [hidden email] mailing list > https://stat.ethz.ch/mailman/listinfo/r-help> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html> and provide commented, minimal, self-contained, reproducible code. ______________________________________________ [hidden email] mailing list https://stat.ethz.ch/mailman/listinfo/r-helpPLEASE do read the posting guide http://www.R-project.org/posting-guide.htmland provide commented, minimal, self-contained, reproducible code.
Open this post in threaded view
|

## Re: Question of "Quantile Regression for Longitudinal Data"

 I've used the function rq.fit.sfn  and rq.fit.panel to estimate a quantile regression on a panel data set.Now I would like to compute an statistic to measure the goodness of fit of this model (T-statistics and   P-value ).  Does someone know how could I do that? best wish
Open this post in threaded view
|

## Re: Question of "Quantile Regression for Longitudinal Data"

 I've used the function rq.fit.sfn  and rq.fit.panel to estimate a quantile regression on a panel data set.Now I would like to compute an statistic to measure the goodness of fit of this model (T-statistics and   P-value ).  Does someone know how could I do that? For formal inference you are better off using rqss() in the quantreg package, but beware that formal inference for shrinkage estimators is still an active research topic. For goodness of fit statistics like the usual regression R^2, see FAQ() item 4 in the quantreg package. url:    www.econ.uiuc.edu/~roger            Roger Koenker email    rkoenker@uiuc.edu            Department of Economics vox:     217-333-4558                University of Illinois fax:       217-244-6678                Urbana, IL 61801
Open this post in threaded view
|

## Re: Question of "Quantile Regression for Longitudinal Data"

 Thanks for your help RKoenker I want to deal with the problem through  bootstrap.so I can get p-value and T-statistics. Do you think so?
Open this post in threaded view
|

## Re: Question of "Quantile Regression for Longitudinal Data"

 How to interprete the results of panel data models of R? I estimate a adapted form of Koenker's (2004) suggestion for a quantile regression approach with panel data, for my data:   rq.fit.panel <- function(X,Y,s,w,taus,lambda)    { require(SparseM)     require(quantreg)  K <- length(w) if(K != length(taus)) stop("length of w and taus must match")  X <- as.matrix(X)      p <- ncol(X)      n <- length(levels(as.factor(s)))      N <- length(y) if(N != length(s) || N != nrow(X)) stop("dimensions of y,X,s must match")      Z <- as.matrix.csr(model.matrix(~as.factor(s)-1))      Fidelity <- cbind(as(w,"matrix.diag.csr") %x% X,w %x% Z)      Penalty <- cbind(as.matrix.csr(0,n,K*p),lambda*as(n,"matrix.diag.csr"))      D <- rbind(Fidelity,Penalty)      y <- c(w %x% y,rep(0,n))  a <- c((w*(1-taus)) %x% (t(X)%*%rep(1,N)),  sum(w*(1-taus)) * (t(Z) %*% rep(1,N)) + lambda * rep(1,n))  rq.fit.sfn(D,y,rhs=a)  }         But i don't have the estimate significances, i don't know identify the results below: \$coef  [1]  1.02281339 -0.18750668 -0.13688807 -0.04180458 -0.01367417  1.02872440 -0.18055062 -0.13003224 -0.03829135 -0.01409369  1.03377335 -0.16649845 -0.11669812 [14] -0.03854060 -0.01438620  1.03851101 -0.15328087 -0.10440359 -0.03871744 -0.01465492  1.04330584 -0.14660960 -0.09670756 -0.03465501 -0.01430647 -0.29187982 [27] -0.21831160 -0.11295134 -0.21530494 -0.15664777 -0.13840296 -0.03224749 -0.11692122 -0.11237144 -0.15112171 -0.10385352 -0.08385934 -0.16090525 -0.30349309 [40] -0.16121494 -0.03106264 -0.16299994 -0.03182579 -0.22271685 -0.08251486 -0.29031224 -0.19680023 -0.20004209 -0.05601186 -0.21140762 -0.04254752 -0.01864703 \$ierr [1] 0 \$it [1] 16 \$time [1] 0 ##summary rq  summary(rq)      Length Class  Mode   coef 52     -none- numeric ierr  1     -none- numeric it    1     -none- numeric time  1     -none- numeric Anybody help me?