Skip to content

Nonnormal Residuals and GAMs

5 messages · David Winsemius, Collin Lynch, COLLINL at pitt.edu

#
Greetings, My question is more algorithmic than prectical.  What I am
trying to determine is, are the GAM algorithms used in the mgcv package
affected by nonnormally-distributed residuals?

As I understand the theory of linear models the Gauss-Markov theorem
guarantees that least-squares regression is optimal over all unbiased
estimators iff the data meet the conditions linearity, homoscedasticity,
independence, and normally-distributed residuals.  Absent the last
requirement it is optimal but only over unbiased linear estimators.

What I am trying to determine is whether or not it is necessary to check
for normally-distributed errors in a GAM from mgcv.  I know that the
unsmoothed terms, if any, will be fitted by ordinary least-squares but I
am unsure whether the default Penalized Iteratively Reweighted Least
Squares method used in the package is also based upon this assumption or
falls under any analogue to the Gauss-Markov Theorem.

Thank you in advance for any help.

	Sincrely,
	Collin Lynch.
#
On Nov 6, 2013, at 12:46 PM, Collin Lynch wrote:

            
The default functional link for mgcv::gam is "log", so I doubt that your theoretical understanding applies to GAM's in general. When Simon Wood wrote his book on GAMs his first chapter was on linear models, his second chapter was on generalized lienar models at which point he had written over 100 pages, and only then did he "introduce" GAMs. I think you need to follow the same progression, and this forum is not the correct one for statistics education. Perhaps pose your follow-up questions to CrossValidated.com
#
your theoretical understanding applies to GAM's in general. When Simon
 Wood wrote his book on GAMs his first chapter was on linear models, his
 second chapter was on generalized lienar models at which point he had
 written over 100 pages, and only then did he "introduce" GAMs. I think
 you need to follow the same progression, and this forum is not the
 correct one for statistics education. Perhaps pose your follow-up
 questions to CrossValidated.com

David, thank you for your advice, has the default changed for mgcv::gam?
Based upon the help pages for the version I have (1.7-27) I had thought
that the default family was gaussian() with link "identity".

In any event I will look again at Simon Woods' book and consider
CrossValidated in the future.

	Best,
	Collin.
#
On Nov 6, 2013, at 5:44 PM, Collin Lynch wrote:

            
I may have gotten this wrong by only referring to my memory. I'm not able to tell by looking at either         ?mgcv::gam or ?gam::gam pages where I picked up this notion.
#
Ok, thanks.