multivariate ordered probit specification
Posted: Mon Apr 18, 2011 8:51 pm
Hi all,
I am an undergraduate and am trying to get the maximum likelihood estimates for a multivariate ordered probit regression. I know in the binary case where the dependent variable takes the value of 0,1 that using logl this is specified as:
@logl logf
param c(1)
mue = @cnorm(c(1))
logf = dependent*log(mue) + (1-dependent)*log(1-mue)
However if this is to be expanded to say 3 possibilities, with the dendent variable take the values 0,1, 2 how is the model specified using logl as opposed to just the estimate equation?
Thanks!
I am an undergraduate and am trying to get the maximum likelihood estimates for a multivariate ordered probit regression. I know in the binary case where the dependent variable takes the value of 0,1 that using logl this is specified as:
@logl logf
param c(1)
mue = @cnorm(c(1))
logf = dependent*log(mue) + (1-dependent)*log(1-mue)
However if this is to be expanded to say 3 possibilities, with the dendent variable take the values 0,1, 2 how is the model specified using logl as opposed to just the estimate equation?
Thanks!