Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, is already up and running.

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

RE: st: Need help on using external global in st_view to avoid recompiling in Mata

From   <>
To   <>
Subject   RE: st: Need help on using external global in st_view to avoid recompiling in Mata
Date   Fri, 30 Mar 2012 13:25:03 -0400

Hello Nick,

Many thanks for taking the time to look at my problem.

I have to admit, however, that I am not sure to fully understand what you suggest (sorry).  

I am slowly changing the way I do things to address problem in Mata, but it's 
not always easy as old (and sometime bad) programming habits are deeply rooted 
in me ;o)

For example, I use to write:



U_0  = beta_y:*y_0 :+ beta_h:*h_0
U_20  = beta_y:*y_20 :+ beta_h:*h_20
U_40  = beta_y:*y_40 :+ beta_h:*h_40

sum_eU = exp(U_0) :+ exp(U_20) :+ exp(U_40)

but working with matrices such as 


reduces the number of arguments and intermediate step and allow me to work in a 
more generic way.  It makes it easier to change the number of labour supply choice and
it also gives me the same results (hopefully not because of luck).   

So far, I have compiled a few versions of my function (with different names) where 
the only things that vary is the number (and name) of variables that are used in 
st_view (e.g. different choices of hours).  Of course, I use different function
on different data set that have been constructed appropriately.  If I understand
correctly, this is the way to do it.

Many thanks,
Jean-François Bertrand

-----Original Message-----
From: [] On Behalf Of Nick Cox
Sent: March 30, 2012 11:56 AM
Subject: Re: st: Need help on using external global in st_view to avoid recompiling in Mata

I would not think this way. Write your code to take what varies as an input argument.


On Fri, Mar 30, 2012 at 4:32 PM,  <> wrote:

> I have written a Mata function to use with -moptimize- or -ml- to 
> estimate a discrete labour supply model.  The simplified example below 
> works for data where I would have 3 distinct hours choice (e.g. 0 
> hours, 20 hours and 40 hours) with income for each choice.
> ----
> version 11.2
> mata:
> function LSmodel_d0(transmorphic scalar M, real scalar todo, real 
> rowvector b, real scalar lnf, real rowvector g, real matrix H) {
>        real colvector beta_y, beta_h
>        real colvector U
>        real colvector sum_eU
>        real colvector y_chosen
>        real colvector h_chosen
>        real colvector L_1
>        real matrix y_i
>        real matrix h_i
>        beta_y          = moptimize_util_xb(M,b,1)
>        beta_h          = moptimize_util_xb(M,b,2)
>        st_view(y=.,.,"y_chosen")
>        st_view(h=.,.,"h_chosen")
>        st_view(y_i=.,.,("y0","y20","y40"))
>        st_view(h_i=.,.,("h0","h20","h40"))
>        U  = beta_y:*y_chosen :+ beta_h:*h_chosen
>        U_i  = beta_y:*y_i :+ beta_h:*h_i
>        sum_eU = quadrowsum(exp(U_i))
>        L_1 = (U :- ln(sum_eU))
>        lnf = moptimize_util_sum(M,L_1) }
> mata mosave mata_LS_singles_quad_d2(), dir(PERSONAL) replace
> end
> ----
> Since I might use different hours choices for different group of the 
> population (as done in the economic literature), I would like to avoid 
> having to recompile the code if I want to run my estimator on, for 
> example, 5 distinct hours choice (e.g. 0 hours, 10 hours, 20 hours, 30 hours and 40 hours) so that:
> st_view(y_i=.,.,("y0","y10","y20","y30","y40"))
> st_view(h_i=.,.,("h0","h10","h20","h30","h40"))
> I was thinking of defining a global macro that could vary depending on the data I use.
> For example, I could use:
> global h_i = `""h0","h20","h40""'
> or
> global h_i = `""h0","h10","h20","h30","h40""'
> depending on the data used.
> Is it possible?
> My attempts to use pointers and global macro (-findexternal()- or 
> -valofexternal()-) have not worked so far.

*   For searches and help try:

*   For searches and help try:

© Copyright 1996–2015 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index