[R] R vglm new family writing: mix Poisson/multinomial

Benoit J benoit_jottreau at yahoo.fr
Mon Jan 7 10:53:50 CET 2008

Hi dear R users,

I would like to know if there is a simple way to define a vglm family which
would be a mix of poisson variables and bernoulli variables (0/1 response)
for idea this would be invoked like this:


where the n's give the number of each type of response.


and a simpler question : How to use constraints in rrvglm?

for example, i want to regress X1 X2 X3 X4 (poisson variables with log-link)
on two factor-type variables N1, N2 but i would like to use X1 and X2 [or
something like 1(X1>X2)] in the estimation of X3 and X4 (but not in X1 X2
estimation of course).

i can do it with simple vglm (with the constraints matrices) but with
rr-vglm i can't.

Excuse me if my questions are silly and for my bad english.

Benoit J.

PS: the final purpose would be to regress something like 
X1,X2,X3,X4 1(X1>X2) 1(X3>X4) on N1, N2  (for the part depending on X1,X2)
and N1,N2,X1,X2 [or N1,N2,1(X1>X2) ] for the responses depending on X3,X4.

X3,X4 could be interpreted like second period variables and X1,X2 like first
period variables.

View this message in context: http://www.nabble.com/R-vglm-new-family-writing%3A-mix-Poisson-multinomial-tp14660710p14660710.html
Sent from the R help mailing list archive at Nabble.com.

More information about the R-help mailing list