R help - Mar 2009 - latex output of regressions with standardized regression coefficients and t-statistics based on Huber-White

Hello,

first of all: I'm new to R and have only used SPSS befor this (which 
can't do this at all...).

I'm trying to output some regression results to latex. The regressions 
are normal OLS and I'm trying to output the results with standardized 
regression coefficients and t-statistics based on "Huber-White sandwich 
estimator for variance". The final result should be one big table for 
several regressions with this data (and N, R^2).

I found the mtable package, which does what I need for lm(...) 
regressions, lm.beta to output standardized coefficients and package 
"sandwich" to output the corrected t-statistics (using
"coeftest(x, vcov
= vcovHAC(x))").

What I couldn't figure out how to put that all together, so that I get a 
latex table with standardized regression coefficients and corrected 
t-statistics based on H-W sandwich.

Is there a way to get this in R and if yes: how?

Thanks in advance,

Jan
-- 
jasc at gmx.net

Jan Schulz wrote:
> Hello,
> 
> first of all: I'm new to R and have only used SPSS befor this (which 
> can't do this at all...).
> 
> I'm trying to output some regression results to latex. The regressions 
> are normal OLS and I'm trying to output the results with standardized 
> regression coefficients and t-statistics based on "Huber-White
sandwich
> estimator for variance". The final result should be one big table for 
> several regressions with this data (and N, R^2).
> 
> I found the mtable package, which does what I need for lm(...) 
> regressions, lm.beta to output standardized coefficients and package 
> "sandwich" to output the corrected t-statistics (using
"coeftest(x, vcov
> = vcovHAC(x))").
> 
> What I couldn't figure out how to put that all together, so that I get
a
> latex table with standardized regression coefficients and corrected 
> t-statistics based on H-W sandwich.
> 
> Is there a way to get this in R and if yes: how?
> 
> Thanks in advance,
> 
> Jan
Standardized regression coefficients are getting to be out of fashion 
for a host of reasons including the emphasis they place on binary 
predictors of very low prevalence.

Frank
-- 
Frank E Harrell Jr   Professor and Chair           School of Medicine
                      Department of Biostatistics   Vanderbilt University

Hello again,

Jan Schulz schrieb:
> What I couldn't figure out how to put that all together, so that I get
a
> latex table with standardized regression coefficients and corrected 
> t-statistics based on H-W sandwich.
Seems that there is a way, I just can't figure out how :-)

It seems to me, that I have to define a costume 'getSummary' function, 
which will return a costum summary which in my case has the lm.beta as 
coefficients and similar the t-statistics based on the vcovHAC().

Unfortunately I did not find any examples how to define my own 
getSummary and after staring for about three hours on the output of 
names(...) and the code of lm.beta, mtable and so on, I gave up.

So first: am I on the right way and if yes: is there a example of a 
costume getSummary function and how to actualy call my mtable with that 
costume function?

Bonus question: how can I change the formating of the output to have my 
t-statistics next to the betas, not under it?

Thanks in advance,

Jan
-- 
jasc at gmx.net