Page 1 of 1
Cross and Direct Elasticity calculation for Nested Logit
Posted: 06 Apr 2022, 06:39
by lucia
Hi Professor,
I have estimated a nested logit model using Apollo. I wonder how I could calculate the Direct Elasticity. Is that possible to provide an example in calculating it in R? I think for Cross-Elasticity within the nest is the same with Multinominal Logit, how about cross-elasticity between nests?
Thank you!
Re: Cross and Direct Elasticity calculation for Nested Logit
Posted: 06 Apr 2022, 07:56
by stephanehess
Hi
other than calculating them analytically, you can approximate them numerically just as in the MNL example MNL_SP_covariates.r
Stephane
Re: Cross and Direct Elasticity calculation for Nested Logit
Posted: 06 Apr 2022, 08:24
by lucia
Hi Professor Hess,
Thank you for your prompt reply!
I have another question relating to arc elasticity for dummy variables.
I manually changed the variable to 0 and to 1 by adding two new columns, but when I calculate the arc elasticity, it shows an Error in predictions_base[, "auto"]: incorrect number of dimensions. I am not sure whether I calculate it in the wrong way.
I attached the code and description below for you to review.
#add two colume
database$rail_use0 <- c(0:0)
database$rail_use1 <- c(1:1)
#estimate 0 model
database$rail_use = database$rail_use0
apollo_inputs = apollo_validateInputs()
predictions_base = apollo_prediction(NL_model, apollo_probabilities, apollo_inputs, prediction_settings=list(runs=30))
#estimate 1 model
database$rail_use = database$rail_use1
apollo_inputs = apollo_validateInputs()
predictions_new = apollo_prediction(EC_model, apollo_probabilities, apollo_inputs)
#elasticity calculation
(predictions_new[,"rail"]-predictions_base[,"rail"])/(predictions_new[,"rail"]+predictions_base[,"rail"])
Thank you!
Lucia
Re: Cross and Direct Elasticity calculation for Nested Logit
Posted: 06 Apr 2022, 14:03
by stephanehess
Lucia
first, your denominator doesn't look correct, as you should only have the base prediction in, not the sum of base and future.
I believe the error you are getting is because for your base prediction, you are using sampling from the asymptotic variance covariance matrix, so try without runs
Stephane
Re: Cross and Direct Elasticity calculation for Nested Logit
Posted: 08 Apr 2022, 06:05
by lucia
Hi Prof. Hess,
Thank you very much for your clarification, really appreciate it!
Kind Regards,
Lucia