I am fitting a multinomial logit model with ridge penalty and in turn estimating the value of time (VOT) or availability to pay (WTP). I want to work with real and simulated data. For the real data I have 3 modes of transport: car, bus and train; where each individual or observation has its travel time and its cost for each mode of transport, that is, i want estimate $\beta_{0}^{auto},\beta_{time}^{auto},\beta_{cost}^{auto},\beta_{0}^{bus},\beta_{time}^{bus},\beta_{cost}^{bus}$.

My question is, how do I choose the lambda penalty parameter? if I want to verify if there is indeed a reduction in the variance of the VOT estimators for the "car" ($var(\hat{VOT}_{auto})$) mode and the VOT for the "bus" ($var(\hat{VOT}_{bus}$) mode, if in the model I consider the train as a reference category? the variance of the VOT estimators I want to determine through the Delta method.

I choose the parameter according to the MSE (mean square error) of the $\beta$ estimators ($MSE(\hat{\beta})=variance+bias^2$)? or by the prediction error using the same data with which I adjusted the multinomial logit model with ridge penalty? or is there a better method or idea than what I propose?

I will read them carefully, thank you very much!


Your Answer

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

Browse other questions tagged or ask your own question.