Nonlinear Panel Regression in R

#1
Hello World.

I have a micro data set for my local CPI basket structured as (panel data with varying monthly periods over set products) in Eviews.
I however, need to run a Logit model on a number of covariates which include a set of dummies. Eviews is limited in this regard (in that it cannot run the logit regression for the panel data). I gather that I could do this in STATA but I would prefer to use a software that I would not have to pay in excess of $1500 especially considering my time constraint (a short one) and need to find a solution now.

So is there any package in "R" that will assist me in doing the relevant analysis. Note. I am by job description an Economist, but by no means an Econometician. Please help.

I have been browsing the list of CRAN Packages as shown in the link below but not sure which may be appropriate :
http://cran.r-project.org/web/packages/available_packages_by_name.html

Kindest regards everyone!
James
 
Last edited:

bryangoodrich

Probably A Mammal
#2
What exactly is the model you are trying to do? What is your response variable? I wouldn't think the consumer price index would be a logit model. In any case, glm handles various models, including logistic regressions. If you need something a bit more diverse, you can define the non-linear function with the nlm package. You'll have to become familiar with the specifics of these functions, though, to get it to do what you want.
 
#3
I am evaluating pricing behaviours among price setters (firms) and evidence of rigidity among various products and other aggregate classifications (eg energy, service, processed foods etc) within the basket of goods. One of the tests to be conducted is to establish the likelihood of a price change(1=y/0=n) given the various covariates. The recommeded model resembles the following where pricecg(+/-) represents a dummy for price change. The tests will be conducted on price increase and decrease separately also.

pricecg(+/-) = sizeprice(+) sizeprice(-) season(m1,m2,m3....) processedfood(y/n) durablegoods(y/n) service(y/n) roundprices(y/n) attractiveprice(y/n) regulated(y/n) duration(mths) stateshock(y/n) competition(#firms)

I have an unbalanced panel dataset ranging from 1995M1to 2006M12 for each CPI micro classification.

I however came across the package pglm and did a test run on the package data. It provides the binomial(logit) options when a panel data framework is supplied. Any other recommendations will be greatly appreciated. Especially as it relates to testing and what likely issues pertaining to data and model I might have to take note and account of before presenting my results.
 
Last edited: