I am hoping that someone on the Zelig mailing list might be able to help
me with the following question. I had previously posted it on the R-help
list but didn't get any replies.
I need to estimate a model that predicts the proportional split of
travel among the vehicles of a household based on vehicle
characteristics such as age, fuel economy, and travel cost per mile. The
model estimation dataset has a record for each household vehicle with
information about the vehicle, the household, and the proportion of the
total household vehicle travel using that vehicle. I have not been able
to figure out how use multinomial logit estimation functions in R to
predict proportions rather than categorical probabilities. I have found
from searching the web that there is a Stata function, FMLOGIT, that
will do what I want. Does anyone know how this can be done in R? All of
my model estimation scripts for the large model I'm building are in R
and I would like to keep it that way to create a nice replicable set of
scripts and data to document the model. Thanks much.
Here's some additional information. I am currently using linear
regression to estimate the model. Then in application I scale the
proportions coming from the model for the vehicles in each household by
dividing by their sum. Perhaps that is good enough.
Brian Gregor
Senior Transportation Analyst
Oregon Department of Transportation
Transportation Planning Analysis Unit
555 13th Street NE
Salem, OR 97301
503-986-4120
Show replies by date