MNP.ox

Multinomial Probit using FiveO.

This is an example of how to implement an econometric objective. FiveO is not designed to handle sample selection and other standard tasks for the user. However, built in features of Ox are readily available for those purposes.

xGHKMNP : xMNP

Public fields
ghk	const
sigfree	const
SigLT	const
Public methods
SetGHK
vfunc		Compute and return the vector of log-likelihoods at the current parameters.
xGHKMNP		GHK based objective for MNP log likelihood.
Enumerations
Anonymous enum 1

Inherited methods from xMNP:: Estimate, SetD, xMNP

Inherited fields from xMNP:: betas, D, indY, J, Jvals, lk, namearray, NN, nX, X, Y

xGQMNP : xMNP

Public fields
Npts	const	# of nodes in quadrature
Public methods
vfunc		Compute and return the vector of log-likelihoods at the current parameters.
xGQMNP		Gauss-Hermite based objective for MNP log likelihood.

Inherited methods from xMNP:: Estimate, SetD, xMNP

Inherited fields from xMNP:: betas, D, indY, J, Jvals, lk, namearray, NN, nX, X, Y

xMNP

Public fields
betas	const	Array of J-1 parameter blocks, one for each equation except Y=0.
D
indY	const	N x J matrix of permutations the first column is index of Y second column is first non-Y choice third column is second non-Y choice etc.
J	const	J : number of options
Jvals	const	integer codes of choices
lk
namearray	const	Labels for variables
NN	const	indexing vector
nX	const	K : number of X variables
X	const	X : matrix of exog variables
Y	const	Y : discrete endog vector
Public methods
Estimate		Estimate the model using BHHH, print results.
SetD
xMNP		Create a new MNP model.

xGHKMNP

Enumerations
Anonymous enum 1		identity, onlydiag, lowertriangle, SigmaOptions

ghk

const decl ghk [public]

SetGHK

sigfree

const decl sigfree [public]

SigLT

const decl SigLT [public]

vfunc

xGHKMNP :: vfunc ( )

Compute and return the vector of log-likelihoods at the current parameters.

xGHKMNP

xGHKMNP :: xGHKMNP ( R , iSigma , fn , Yname , Xnames )

GHK based objective for MNP log likelihood.

Parameters:

iSigma

Yname

Xnames

xGQMNP

Npts

const decl Npts [public]

# of nodes in quadrature

vfunc

xGQMNP :: vfunc ( )

Compute and return the vector of log-likelihoods at the current parameters.

xGQMNP

xGQMNP :: xGQMNP ( Npts , fn , Yname , Xnames )

Gauss-Hermite based objective for MNP log likelihood.

Parameters:

Npts

Yname

Xnames

xMNP

betas

const decl betas [public]

Array of J-1 parameter blocks, one for each equation except Y=0.

D

decl D [public]

Estimate

xMNP :: Estimate ( )

Estimate the model using BHHH, print results.

indY

const decl indY [public]

N x J matrix of permutations
the first column is index of Y
second column is first non-Y choice
third column is second non-Y choice
etc.

J

const decl J [public]

J : number of options

Jvals

const decl Jvals [public]

integer codes of choices

lk

decl lk [public]

namearray

const decl namearray [public]

Labels for variables

NN

const decl NN [public]

indexing vector

nX

const decl nX [public]

K : number of X variables

SetD

xMNP :: SetD ( )

X

const decl X [public]

X : matrix of exog variables

xMNP

xMNP :: xMNP ( fn , Yname , Xnames )

Create a new MNP model.

Parameters:

fn	string, a file to load the data from using Ox `Database.Load()`
Yname	string, name or label of the column in the file that contains `Y` Yname can contain any integers. MNP will translate the unique sorted values into 0...Jvals-1
Xnames	a string of the form var1 var2 ... varN

Comments:

Observations with any missing data are deleted.
A constant column is appended at the end of the X matrix as in Stata.
Summary statistics are reported.

Y

const decl Y [public]

Y : discrete endog vector