Re: st: which STATA version for Poi's QUAIDS command

["Brian P. Poi" <bpoi@stata.com>]

On 03/18/2013 04:15 PM, Luca Tiberti wrote:
> Hi Nick, thanks for this
>
> here is the output of describe, short
>
>
> obs:         3,860
> vars:            39                          18 Mar 2013 15:37
> size:       455,480
> Sorted by:  serial
>       Note:  dataset has changed since last saved
>
> For your info, I had already tried by keeping only variables finally
> entering into the demand system and compress them. However I was not
> able to run succesfully the command.
>
> Thanks for your help,
> Luca
Luca,

How many goods are in your demand system?  How many demographic variables are you specifying?

To estimate the model, -quaids- must create temporary variables to hold observation-level derivatives of each equation with respect to each of the parameters.  The details are in the methods and formulas section of [R] nlsur.

For a basic AIDS model, with three goods there are 7 free parameters and 2 estimated equations, for a total of 14 temporary variables -quaids- needs to create.  With four goods, there are 12 free parameters and 3 estimated equations, for a total of 36 temporary variables.  If my math is correct, a basic AIDS model with 15 goods has 133 parameters and 14 estimated equations for a total of 1,862 temporary variables.  If you have 16 goods, there are 2,250 derivatives that must be stored in temporary variables, too many for Stata/IC to handle.

If you include the quadratic income term (the QUAIDS model), then you will hit Stata/IC's variable limit if you try to fit a 15-good model.

Throw one demographic variable into the QUAIDS model, and you are down to 13 goods.  Five demographic variables and you're down to 11 goods.

In short, having many goods and demographics can quickly balloon the number of parameters in the model and hence the number of temporary variables needed during estimation.

I hope this helps.

   -- Brian Poi
   -- bpoi@stata.com



*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/