mlogit {mlogit} | R Documentation |
Estimation by maximum likelihood of the multinomial logit model, with choice-specific and/or individual specific variables.
mlogit(formula, data, weights = NULL, ...) ## S3 method for class 'mlogit': print(x, digits = max(3, getOption("digits") - 2), width = getOption("width"), ...) ## S3 method for class 'mlogit': summary(object, ...) ## S3 method for class 'summary.mlogit': print(x, digits = max(3, getOption("digits") - 2), width = getOption("width"), ...) ## S3 method for class 'mlogit': print(x, digits = max(3, getOption("digits") - 2), width = getOption("width"), ...) ## S3 method for class 'mlogit': logLik(object, ...) ## S3 method for class 'mlogit': vcov(object, ...)
x, object |
an object of class mlogit |
formula |
a symbolic description of the model to be estimated, |
data |
the data, |
weights |
an optional vector of weights, |
digits |
the number of digits, |
width |
the width of the printing, |
... |
further arguments. |
Let J
being the number of alternatives. The formula may
include choice-specific and individual specific variables. For the
latter, J-1
coefficients are estimated for each
variable. Choice and individual specific variables are separated by a
|
. For example, if x1
and x2
are choice specific
and z1
and z2
are individual specific, the formula
y~x1+x2|z1+z2
describe a model with one coefficient for
x1
and x2
and J-1
coefficients for z1
and
z2
. J-1
intercepts are also estimated. A model without
intercepts is defined by the formula : y~x1+x2-1|z1+z2
. To
obtain alternative specific coefficients for the choice-specific
variable x2
, use : y~x1+x2+x2:alt|z1+z2
(replace
alt
by the relevant variable name if the alternative index is
provided. Models with only choice-specific or individual-specific
variables are respectively estimated by y~x1+x2
and
y~1|z1+z2
.
The model is estimated with the maxLik
package and the
Newton-Raphson method, using analytic gradient and hessian.
An object of class "mlogit"
, a list with elements
coefficients
, logLik
, hessian
,
gradient
, call
, est.stat
,
residuals
and fitted.values
.
Yves Croissant
McFadden, D. (1973) Conditional logit analysis of qualitative choice behavior, in P. Zarembka ed., Frontiers in Econometrics, New-York: Academic Press.
McFadden, D. (1974) ``The Measurement of Urban Travel Demand'', Journal of Public Economics, 3, pp. 303-328.
Train, K. (2004) Discrete Choice Modelling, whith Simulations, Cambridge University Press.
mlogit.data
to shape the data. multinom
from package nnet
performs the estimation of the multinomial
logit model with individual specific variables
# Heating data, from the Ecdat package data("Heating",package="Ecdat") # Heating is a "horizontal" data.frame with three choice-specific # variables (ic: investment cost, oc: operating cost) and some # individual-specific variables (income, region, rooms) Heatingh <- mlogit.data(Heating,cvar=c(ic=3,oc=8,pb=17), shape="hor.var",choice="depvar") # a model with two choice-specific variables summary(mlogit(depvar~ic+oc,data=Heatingh)) # same wihtout intercept summary(mlogit(depvar~ic+oc-1,data=Heatingh)) # a model with choice-specific and individual-specific variables summary(mlogit(depvar~ic+oc|income+rooms,data=Heatingh)) # a model with choice-specific coefficients for a choice-specific variable summary(mlogit(depvar~ic+oc+oc:alt,data=Heatingh)) # a model with only individual-specific variables summary(mlogit(depvar~1|income+rooms,data=Heatingh)) # the same model estimated with multinom from the nnet package library(nnet) summary(multinom(depvar~income+rooms,data=Heating))