Differences Between LIMDEP and NLOGIT
NLOGIT Version 6 is an extension of LIMDEP 11 that provides programs for estimation, model simulation and analysis of multinomial choice data, such as brand choice, transportation mode, and all manner of survey and market data in which consumers choose among a set of competing alternatives.
NLOGIT 6 includes all the features and capabilities of LIMDEP 11 including data handling, estimation, matrix, algebra, and so on - plus NLOGIT’s FIML estimation programs. (NLOGIT 6 may not be added to a LIMDEP 11 license at a later date.)
The differences between LIMDEP 11 and NLOGIT 6 are listed below:
LIMDEP 11 contains several of the basic forms of the discrete choice models that appear in NLOGIT 6 as well. LIMDEP 11 includes all of the following models for discrete choice:
- All forms of the probit, logit and other binary choice models.
- All forms of the ordered choice models.
- Bivariate probit and all variants, such as sample selection and partial observability.
- Multivariate probit models. Note, this is not the same as multinomial probit. Multivariate probit refers to a multiple equations system of probit equations. Multinomial probit refers to a multivariate normally distributed system of utility functions underlying a multinomial choice setting.
- Panel data forms of all the models listed above, including fixed and random effects, random parameters, and latent class models.
- The basic (McFadden style) conditional multinomial) logit model. This is the basic discrete choice model based on the type 1 extreme value distribution. This is the standard model documented, for example, in modern econometrics texts such as Greene (2017). The two LIMDEP commands for this model are MLOGIT, which is for models that are based on individual characteristics, such as age, sex, etc., and CLOGIT, which is generally based on choice attributes. (CLOGIT allows mixtures of characteristics and attributes.)
NOTE: The CLOGIT and MLOGIT forms are the basic platforms for the extensions in NLOGIT.
The following extensions of the discrete choice models appear only in NLOGIT 6:
- Nested logit models: Estimated by full information maximum likelihood or two step maximum likelihood. This includes the generalized nested logit model which allows the branches to share alternatives (probabilistically).
- Covariance heterogeneity - this is an extension of the nested logit model to allow heteroscedasticity in the utility functions.
- Heteroscedastic extreme value.
- Random parameters logit. LIMDEP contains a random parameters version of the simple binomial logit for a single 0/1 dependent variable. The RPL model in NLOGIT extends this to the CLOGIT model for discrete choice among several alternatives.
- Generalized mixed logit and special cases such as scaled multinomial logit.
- Mixed logit with nonlinear utility functions.
- Error components logit (with the random parameters model or by itself).
- Multinomial probit and multiperiod multinomial probit.
- Latent class multinomial logit model.
- NLOGIT also contains the model simulator for all of the extensions noted.