I am at a loss as to which R package and approach to use for my modeling. The background is: I have a transportation survey dataset combined with many other GIS datasets. I am attempting to determine the effects of household and urban characteristics surrounding each respondent on their choice to drive or take any other form of transportation.

Having read the Bivand book "Applied Spatial Data Analysis", and most of the documentation of the packages listed below, I am still unclear as to whether this should be done as a point pattern model, since I have a point for each location of survey respondents, or whether it is a spatial probit model, etc. Much of the documentation has only function definitions with very little examples of application from start to finish, and I can't make heads or tails of which chapter of Bivand would be the correct one to follow. From this, can anyone point me in the direction of whether I should be using R packages spatstat, McSpatial, GWmodel, spatialprobit, spdep, sphet, splancs or some other package? I am also having trouble creating any type of spatial weights matrix other than a binary neighbor weight, but would like to create a Gaussian or some other exponentially declining weighting scheme for spatial dependency.


This looks like a multinomial regression problem to me. I.e, a logistic regression with more than two choices.


choice ~ income, age, zip code, distance to next bus stop, .....

and choice being one of bus / bike / car / walk

Maybe the following posts are helpful:

http://www.jameskeirstead.ca/blog/how-to-multinomial-regression-models-in-r/ https://onlinecourses.science.psu.edu/stat504/node/172

Or search for "multinomial logistic regression", aka as "polytomous logistic regression"

There is even a package in R, "polytomous", but I have no experience with it.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.