Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: st: using lowess to compare groups


From   "Mollon, Josephine" <josephine.j.mollon@kcl.ac.uk>
To   "statalist@hsphsun2.harvard.edu" <statalist@hsphsun2.harvard.edu>
Subject   RE: st: using lowess to compare groups
Date   Mon, 2 Dec 2013 09:48:26 +0000

Thank you for your help, much appreciated! I am new to Stata so will look into what you suggest.

-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Nick Cox
Sent: 28 November 2013 15:25
To: statalist@hsphsun2.harvard.edu
Subject: Re: st: using lowess to compare groups

sprinkled freely here without further flags.

In principle, you could build a model based on lowess smoothing of one or more predictors with or without such smoothing of any other predictors. That way lie generalised additive models, which I don't think are well supported in Stata beyond what is revealed by -search gam-. (-mlowess- (SSC) isn't an exception.)  In 1990 they looked like the future to some of us; in practice statistical modelling has largely moved since in quite different directions.

One key point is that you wouldn't (couldn't) end up with an equation summarizing your model that could be cited in  a paper or thesis.
Communities vary on whether that is a flaw or adherence to an outdated statistical style.

Another even more insidious problem is that -lowess- (Stata command) is just one flavour of lowess (or loess, or locfit) and not necessarily one that can be matched by implementations in other software.

Although I call that a problem, it is only that if you read more into
-lowess- results than they deserve.

In essence, lowess is best treated as an exploratory method that helps signal whether transformations of your predictors or response would be helpful. Having used its results, you can throw them out.

Alternatively, fractional polynomials may offer some or all of what you seek and they certainly are well supported in Stata.
Nick
njcoxstata@gmail.com


On 28 November 2013 15:07, Mollon, Josephine <josephine.j.mollon@kcl.ac.uk> wrote:
> Hi, I am using Stata 10 and want to model the association between age and a continuous variable, but also look at group differences in this association. The groups are symptomatic and control groups. I have used lowess smoothing to create graphs of my data, but don't know how to use lowess to obtain a statistical model.
> Any help is much appreciated!
>
> *
> *   For searches and help try:
> *   http://www.stata.com/help.cgi?search
> *   http://www.stata.com/support/faqs/resources/statalist-faq/
> *   http://www.ats.ucla.edu/stat/stata/

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/


© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index