Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

st: RE: Hessian trigger "too many literals" error with "ml model"

From   <>
To   <>
Subject   st: RE: Hessian trigger "too many literals" error with "ml model"
Date   Wed, 29 Feb 2012 15:33:10 -0500

Hello all,

I work with STATA 11, 64 bit.

I am estimating a discrete labour supply model for single individuals.  My model is similar to what is done in many economic paper on this topic.

When I use a quadratic utility function including fixed cost for working such that:

U = beta_y * income + beta_h * hours + alpha_ysq * income square  + alpha_hsq * hours square + alpha_yh * (hours times income)   

y = Y - gamma_fcw * dummy_if_working - gamma_fcwft * dummy_if_working_40_hours_per_week

Where y = income, h = hours, ysq = income square, hsq = hours square and yh = interaction between hours and income and where Y represent income net of taxes.

Ml model works perfectly with d0, d1 and d2 approach (d2 being much faster and stable).

When I relax the independence of irrelevant alternative of my extreme value distribution by using simulated maximum likelihood approach to include a random term with income for example (I am using Halton generated numbers), d2 stop working as I get a "too many literals" with my Hessian, but d0 and d1 provides me with similar results.

If I omit fixed cost of working variables, d2 works.

Is there a way to increase the number of "literals" as d2 is much faster?


*   For searches and help try:

© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index