[R] Re gression for loop test HELP! URGENT!

Daniel Malter daniel at umd.edu
Tue Jul 21 07:02:50 CEST 2009


Hi, first, your initial statement of what you wanted to do was obviously
ambiguous enough to confuse the responders. Therefore, clarity helps greatly
in getting an accurate response. If I understand correctly, you have run ONE
model on whatever data (also often called testing sample). Now you want to
assess how well this ONE estimate derived from the testing sample predicts
data from five (or any other arbitrary number of) holdout samples.

In order to do that, the example I have provided works perfectly fine. The
only thing you do not do is run multiple regressions in the first place.
Instead you run only one initial regression on the testing sample and
predict into the holdout samples. This, however, is only a slight change of
the procedure I have outlined. The fit to the holdout samples is assessed
with measures that any statistics/econometrics book deals with if it has a
section on prediction. 

Best,
Daniel
-------------------------
cuncta stricte discussurus
-------------------------

-----Ursprüngliche Nachricht-----
Von: r-help-bounces at r-project.org [mailto:r-help-bounces at r-project.org] Im
Auftrag von Rbeginner
Gesendet: Monday, July 20, 2009 10:50 PM
An: r-help at r-project.org
Betreff: Re: [R] Re gression for loop test HELP! URGENT!


I think the problem is that I've been getting replies about how to make new
regressions, when in fact, I need to use the one I've produced already to
fit new data, 5 rows at a time, to see if it is also a good representation
of further data. From the replies, I'm getting the impression that the only
way I can do that is bye producing more regressions and calculating the
error, but I'm not sure how I should do that, if I get hundreds of new
regressions. I'm thinking, in my primitive programming terms, that I should
ask the system to run through the new data 5 rows at a time and produce some
indication of deviation (error) from the original regression, which would
help me decide whether the original regression is is a good representation
of the new data. Does this make sense?

fm <- lm(M ~ D + O + S)            #this is my original regression, and I
need to use this to fit the test data.
test = data.frame(Mtest,Dtest,Otest,Stest)  #data frame of the test data
attach(test)
for (i in 1:1184){
fmtest <- lm(Mtest ~ Dtest + Otest + Stest, subset=(1:5), data=test)
print(summary(fmtest)) }

#this would only produce a long string of summaries. 
My data is in the form of 
          M       D         O        S
1
2
...
1184

Any suggestions? 



Richard Cotton wrote:
> 
>> I'm new to R, and I've sent this message as a non-member, but since 
>> it's pretty urgent, I'm sending it again now I'm on the mailing list 
>> (Thanks Daniel for your suggestion nevertheless).
>> 
>> I have calculated a regression in the form of M ~ D + O + S, and I 
>> would like to take this regression and test it with other samples, 5 
>> sets of
> M, D,
>> O, and S at a time(I actually have 2000 sets, so it's probably not
> efficient
>> to make each a separate set and then index). Since I'll need to test 
>> the regression for 400 groups, I thought a for loop might be 
>> necessary. I've
> put
>> everything into a data frame already. Can anyone tell me how to write
> the
>> code? I'm especially not sure about how to do the for loop.
>> And then how would I calculate the error of how well the test samples
> fit
>> the original regression?
>> This is for my internship, so it's very urgent.
> 
> Take a deep breath, and think calm thoughts.  Take a look at the 
> posting guide (http://www.r-project.org/posting-guide.html) - it has 
> useful ideas on thinking through your problem.  If you can provide 
> some code then we can see what you want more clearly.
> 
> Show us how you've done your regression what form your data is in.  
> Tell us which tests you'd like to do on the samples.
> 
> If you are stuck with for loops, then take a look at section 9.2.2 in 
> the Intro to R guide that comes with R.  (Click Help -> Manuals -> an 
> Introduction to R in RGui.)
> 
> Regards,
> Richie.
> 
> Mathematical Sciences Unit
> HSL
> 
> 
> ----------------------------------------------------------------------
> --
> ATTENTION:
> 
> This message contains privileged and confidential 
> inform...{{dropped:22}}
> 
> ______________________________________________
> R-help at 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.
> 
> 
> -----
> Regards,
> Richie.
> 
> Mathematical Sciences Unit
> HSL
> 

--
View this message in context:
http://www.nabble.com/Regression-for-loop-test-HELP%21-URGENT%21-tp24564236p
24580889.html
Sent from the R help mailing list archive at Nabble.com.

______________________________________________
R-help at 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.




More information about the R-help mailing list