[Limdep Nlogit List] Marginal effects for unlabeled choice experiment

Jacky Kemp kemp.jacky at gmail.com
Sat Apr 11 03:14:55 EST 2009

Hello all,

Below are the commands I am currently using to estimate a nested logit
model for an unlabeled choice experiment:

nlogit ; Lhs=choice
; tree=(ContractA,ContractB),(Nothing)
; choices=ContractA,ContractB,Nothing
; ru2
; effects: ACRES[ContractA,ContractB,Nothing] $

The nested logit above has a tree such that the branch is purchase a
contract or not, and then the choice is if purchase a contract,
contract A or contract B.

In the above code, if I don't specify the "effects" option, it will
give me just one logit estimate for each attribute on the RHS, not by
contract. I just want to estimate the marginal effect of that

My problem is that the output for these commands will provide, for
example, the marginal effect on all 3 alternatives of an increase in
Contract A's "acres" attribute. However, because the contracts are
unlabeled, it would make more sense to obtain the marginal effect on
choosing a generic contract (contractA OR contractB) of an increase in
the contract's "acres" attribute.

Is there a way in LIMDEP to obtain the marginal effects for a generic
contract? Or has anyone dealt this problem by estimating and reporting
the marginal effects for just one of the contracts? Alternatively,
does it make sense to obtain marginal effects for contracts A and B
separately and then report the average? In my case, the marginal
estimates are slightly different for the two contracts A and B.

Thank you for any suggestions,

Jacky Kemp

Master's Degree Student
Department of Environmental and Natural Resource Economics
University of Rhode Island
Kingston, RI
kemp.jacky at gmail.com

More information about the Limdep mailing list