-gwr- is a wrapper for -glm-. I am not clear why there should be any
confusion, as this is explained in its help. So, yes, if you specified
fam(binomial) link(logit)
that would entail a logit model, and you could get -glm-'s flavour of
Poisson regression by appropriate option choices. It would differ from
-poisson- in small details, but
the limitations of geographically weighted regression are a much bigger
issue.
Nick
[email protected]
wangxin (Fiona)
Can GWR deal with Poisson Regression? since My
dependent is a count number.
In the help gwr, there is a script:
gwr flag class unemp, east(east) north(north)
fam(binomial) link(logit)
Does this mean a GW logit regression? I am a little
confused about this. Any advice is welcome.
--- Nick Cox <[email protected]> wrote:
> -gwr- was published in the Stata Technical Bulletin
> in 1998:
>
> STB-46 sg95 . . . . . . . . . . . . . . .
> Geographically weighted
> regression
> (help gwr if installed) . . . . . . . . . .
> . . . . . . . M. S.
> Pearce
> 11/98 pp.20--24; STB Reprints Vol 8,
> pp.189--194
> applies geographically weighted regression
> to a dataset
> containing geographical reference points
>
> I have not used it, but the implication that it is
> not well documented
> is a little misleading. The article above is freely
> available at the
> Stata website, regardless of whether you or your
> institution ever
> subscribed. See
>
> http://www.stata.com/products/stb/journals/stb46.pdf
>
> Nor is there much mystery about st_0_0_1. You can
> look at the code
> within Stata by
>
> . viewsource gwr.ado
>
> and even without any need to be a Stata programmer
> see that this is the
> name of the data file being used to store results.
> (Nowadays a temporary
> file might well be used.)
>
> On your main point, I don't find it especially
> surprising that it is
> slow, as the thrust of the technique is to get an
> idea of spatial
> variability through repeated modelling, in your case
> thousands of times.
> In addition -gwr- could be speeded up a bit, and
> perhaps even a lot, by
> taking advantage of changes in Stata over the last
> decade, but that
> doesn't help you unless someone does it.
>
> Nick
> [email protected]
>
> Helen Rosenkrantz
>
> I am running a geographically weighted regression
> model in Stata with
> 3,019 observations and 5 variables. I am wondering
> if anyone might be
> able to help me since I have found little guidance
> available online
> regarding gwr in stata (aside from some information
> provided by the
> program creator and help function). It has been
> processing for over 24
> hours and seems to be updating an output dataset
> called "st_0_0_1".
> Included in this mystery dataset are numbers for the
> 5 vars and the
> constant.
>
> Here is my syntax: gwr depvar indvars,east( xcoo)
> north( ycoo)
> reps(1000) test saving(eb_gwr) outfile(eb_gwr)
> mcsave(mc_eb)
>
>
> Has anyone had success in using Stata for GWR?? Any
> help would be
> greatly appreciated!
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
