Error using glm with poisson family and identity link

Hi, Peter: 

      What do you do in such situations? 

      Sundar Dorai-Raj and I have extended "glm" concepts to models 
driven by a sum of k independent Poissons, with the a linear model for 
log(defectRate[i]) for each source (i = 1:k).  To handle convergence 
problems, etc., I think we need to use informative Bayes, but we're not 
there yet.  In any context where things are done more than once [which 
covers most human activities], informative Bayes seems sensible. 

      A related question comes with data representing the differences 
between Poisson counts, e.g., with d[i] = X[i]-X[i-1] = the number of 
new defects added between steps i-1 and i in a manufacturing process.  
Most of the time, d[i] is nonnegative.  However, in some cases, it can 
be negative, either because of metrology errors in X[i] or because of 
defect removal between steps i-1 and i. 

      Comments?
      Best Wishes,
      Spencer Graves

Spencer Graves <spencer.graves at pdf.com> writes:

 

Dear Federico:     Why do you use the "identity" link?  That can
produce situations with an average of (-2) Poisson defects per unit,
for example.  That's physical nonsense. 
   

So is _not_ using the identity link when the model is manifestly
additive on the identity scale. E.g. calibrating differential
spectrofluorometry with photon counters recording linear combinations
of intensities at different wavelengths.

I've bumped into similar situations before (binomial(link=identity), I
think it was then) and the glm.fit algorithm could use improvement in
dealing with the parameter constraints in these cases. With the
standard IRLS algorithm, if the maximum is on the boundary, you
basically hit a random point on the boundary and get stuck there with
a search direction pointing out of the valid region.

 

Spencer Graves, PhD, Senior Development Engineer
O:  (408)938-4420;  mobile:  (408)655-4567