R help - Mar 2011 - rqss help in Quantreg

Dear All,

I'm trying to construct confidence interval for an additive quantile
regression
model.

In the quantreg package, vignettes section: Additive Models for Conditional 
Quantiles
http://cran.r-project.org/web/packages/quantreg/index.html

It describes how to construct the intervals, it gives the covariance matrix for 
the full set of parameters, \theta is given by the sandwich formula
V = \tau (1-\tau) (t(X) \phi X)^(-1) (t(X) X)^(-1) (t(X) \phi X)^(-1)
but gives no derivation of this result. 

Does anyone know how to obtain the above results? I need to construct these 
intervals with several adjustments, so I need to understand how "V" is
constructed. Any proofs?

Thanks, any help is greatly appreciated
Man Zhang


      
	[[alternative HTML version deleted]]

On Mar 20, 2011, at 10:24 PM, Man Zhang wrote:
> Dear All,
> 
> I'm trying to construct confidence interval for an additive quantile
regression
> model.
> 
> In the quantreg package, vignettes section: Additive Models for Conditional
> Quantiles
> http://cran.r-project.org/web/packages/quantreg/index.html
> 
> It describes how to construct the intervals, it gives the covariance matrix
for
> the full set of parameters, \theta is given by the sandwich formula
> V = \tau (1-\tau) (t(X) \phi X)^(-1) (t(X) X)^(-1) (t(X) \phi X)^(-1)
> but gives no derivation of this result. 
> 
> Does anyone know how to obtain the above results? I need to construct these
> intervals with several adjustments, so I need to understand how
"V" is
> constructed. Any proofs?
As the author of the aforementioned vignette I suppose that I am meant to feel
chastised by this request.  I do not.  There is a reasonable expectation in
mathematics and most other fields that readers have some responsibility
to familiarize themselves with the related literature, especially that cited in
the work that they are currently reading.

There seems to be a growing, deplorable expectation that learning a new subject
is
like being spoon fed some sort of pablum -- it isn't, sometimes you need to
chew.

I would remind would be R-posters of the admonition of Edward Davenant
(quoted by Patrick Billingsley in his magisterial P&M)

	I would have a man knockt on the head that should write anything in 
	Mathematiques that had been written of before. 


> 
> Thanks, any help is greatly appreciated
> Man Zhang
> 
> 
> 
> 	[[alternative HTML version deleted]]
> 
> ______________________________________________
> R-help@r-project.org 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.

	[[alternative HTML version deleted]]

Dear All, 

I,m trying to fulfill a constraint nonparametric quantile regression
analysis for monthly stock index  and gdp (159 cases of data) using rqss
function of quantreg package. I need to specify that stock prices are
nondecreasing with growing gdp. 

I tried the following simple code

fit1<-rqss(stock~gdp)
fit2<-rqss(stock~qss(gdp,constraint="I")+time)

but R produces an error message

for the firsts line of the code
Error in rqss.fit(X, Y, tau = tau, rhs = rhs, nsubmax = nsubmax, nnzlmax
nnzlmax,  :
object 'rhs' is not found

for the second line of the code
Error in D %*% B : NA/NaN/Inf when calling external function (argument 7)

If I use returns instead of prices, the analysis goes. But I need to regress
prices.
What is wrong in my specification? Are there any restrictions in the rqss
approach?

--
View this message in context:
http://r.789695.n4.nabble.com/rqss-help-in-Quantreg-tp3392770p3557884.html
Sent from the R help mailing list archive at Nabble.com.
	[[alternative HTML version deleted]]