R help - Jul 2011 - Poisson GLM with a logged dependent variable...just asking for trouble?

Dear R-helpers,

I'm using a GLM with poisson errors to model integer count data as a
function of one non-integer covariate.

The model formula is: log(DV) ~ glm(log(IV,10),family=poisson).

I'm getting a warning because the logged DV is no longer an integer.

I have three questions:

1) Can I ignore the warning, or is logging the DV (resulting in
non-integers) a serious violation of the Poisson error structure?

2) If the answer to #1 is "no, don't ignore it, it's serious"
then can
I use a quasipoisson error structure instead (does not  give the same
warning) and if so are there any pitfalls to using the quasipoisson
model? Are there any better alternatives for count data where the
counts must be logged? Or, should I just abandon logging the DV? In
that case, how could I compare the fit of a Poisson model (without
logging the DV) to that of a GLM with normal errors (with a logged
DV). AIC would not be valid because the DVs are different, right?

3) The quasipoisson model doesn't return an AIC value. Why, and is
there anything I can do to calculate AIC manually, that would allow me
to compare this model to other models?

Many thanks in advance for your help!

Cheers, Mark

Dear Mark,

I think you want glm(DV ~ log10(IV), family=poisson)
Note that the poisson family uses the log-link by default. Hence you don't
need to log-transform DV yourself.

Best regards,

Thierry

----------------------------------------------------------------------------
ir. Thierry Onkelinx
Instituut voor natuur- en bosonderzoek
team Biometrie & Kwaliteitszorg
Gaverstraat 4
9500 Geraardsbergen
Belgium

Research Institute for Nature and Forest
team Biometrics & Quality Assurance
Gaverstraat 4
9500 Geraardsbergen
Belgium

tel. + 32 54/436 185
Thierry.Onkelinx at inbo.be
www.inbo.be

To call in the statistician after the experiment is done may be no more than
asking him to perform a post-mortem examination: he may be able to say what the
experiment died of.
~ Sir Ronald Aylmer Fisher

The plural of anecdote is not data.
~ Roger Brinner

The combination of some data and an aching desire for an answer does not ensure
that a reasonable answer can be extracted from a given body of data.
~ John Tukey
> -----Oorspronkelijk bericht-----
> Van: r-help-bounces at r-project.org [mailto:r-help-bounces at
r-project.org]
> Namens Mark Na
> Verzonden: vrijdag 1 juli 2011 23:10
> Aan: r-help at r-project.org
> Onderwerp: [R] Poisson GLM with a logged dependent variable...just asking
for
> trouble?
> 
> Dear R-helpers,
> 
> I'm using a GLM with poisson errors to model integer count data as a
function of
> one non-integer covariate.
> 
> The model formula is: log(DV) ~ glm(log(IV,10),family=poisson).
> 
> I'm getting a warning because the logged DV is no longer an integer.
> 
> I have three questions:
> 
> 1) Can I ignore the warning, or is logging the DV (resulting in
> non-integers) a serious violation of the Poisson error structure?
> 
> 2) If the answer to #1 is "no, don't ignore it, it's
serious" then can I use a
> quasipoisson error structure instead (does not  give the same
> warning) and if so are there any pitfalls to using the quasipoisson model?
Are
> there any better alternatives for count data where the counts must be
logged?
> Or, should I just abandon logging the DV? In that case, how could I compare
the
> fit of a Poisson model (without logging the DV) to that of a GLM with
normal
> errors (with a logged DV). AIC would not be valid because the DVs are
different,
> right?
> 
> 3) The quasipoisson model doesn't return an AIC value. Why, and is
there
> anything I can do to calculate AIC manually, that would allow me to compare
> this model to other models?
> 
> Many thanks in advance for your help!
> 
> Cheers, Mark
> 
> ______________________________________________
> 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.