[Limdep Nlogit List] Many choice occasions and unbalanced panel - estimation failure of RPL

[linda_thunstrom]

Dear Sirs,

I have panel data (unbalanced) on household consumption
of bread, and am estimating a mixed logit model (RPL). My goal is to
estimate the effect on choice probabilities of
different bread types (5 types), of household characteristics (a set of
dummy variables), the price of the bread types and, finally, the effect of
having purchased the bread type on the previous choice occasion.

The number of choice occasions for each household varies between 1 and
100. Oddly enough, I can estimate a RPL-model on a subsample of households
with no more than 24 choice occasions. As soon as I include households
with more than 24 choice occasions, I fail to estimate the model, and get
the following error message:
Error:   805: Initial iterations cannot improve function.Status=3
Function=  .10260381542D+05, at entry,  .10197190943D+05 at exit
Error:  1025: Failed to fit model. See earlier diagnostic.

There seems to be nothing wrong with the variable I use to define the
number of choice occasions per household. Also, the failure of estimation
does not seem to be due to the large spread in number of choice occasions;
if I try to estimate the model based on households with between, say, 40
and 50 choice occasions, the estimation still fails, although the number
of choice occasions is fairly equal over houceholds. It therefore seems
more likely that something happens in the model (increasing
multicollinearity as the number of choice occasions increases, or limited
capacity in the program?). Has someone else had similar problems, or an
idea of what the problem might be?

My current code is (a bit simplified) the following:

NLOGIT  ; Lhs = choice
        ; Choices = white, sweet, brown, fibre, hard
        ; Rpl
        ; Fcn = pris(N), pp(N)
        ; Pds = pds
        ; Pts = 10
        ; Halton
        ; Model: U(white) = pr*price + pp*prevp /
                 U(sweet) = sweet + pr*price + pp*prevp /
                 U(brown) = brown + pr*price + pp*prevp /
                 U(fibre) = fibre + pr*price + pp*prevp /
                 U(hard) = hard + pr*price + pp*prevp $

Where price varies over time and alternatives, prevp = a dummy equal to 1
if the product was purchased on the previous choice occasion, otherwise
zero. Sweet, brown, fibre and hard are alternative specific constants.
Pds is a variable indicating the number of observations over time
associated with each household in the panel (as mentioned above, the panel
is unbalanced, and this number varies from 1 to 100).

I am very grateful for all input, comments and suggestions!

Very best regards,
Linda