MULTINOMIAL LOGIT MODEL
Multinomial Logit
(MNL) model is similar to the Binary Logit model, except that the dependent
variable is in this case will have multiple discrete outcomes, instead of just
2. Some examples are models predicting consumer choice (choose 1 out of 5
brands), models predicting market share ranks etc. The estimation technique is
very similar to the Binary Logit model, except that instead of predicting the
odds of 1 vs. 0, it predicts the odds of the different outcomes vs. a baseline
outcome. For e.g. for a model with 3 outcomes A, B, C, it estimates odds of B
vs. A and C vs. A.
Multinomial Logit models
are used in applications in marketing that have several distinct outcomes. One
common application is to predict what product or brand a customer is going to
choose. Another application is to predict purchase when the consumer may have
more than two options, e.g. pay a loan installment, pay off and close the loan
or default.


© Copyright 2008
eNumerys Global LLC. All Rights Reserved.