st: RE: RE: RE: RE: RE: query regarding mata

["Nick Cox" <n.j.cox@durham.ac.uk>]

I see. You changed more of the code than I appreciated on a quick
glance. 

But I think your problem is of the same form: you're expecting Mata to
assign v[1,1] before you've told it how big -v- is. 

You're better off initialising -v- too, say to J(1,5,.), _before_ the
loop. 

Nick 
n.j.cox@durham.ac.uk 

Abhimanyu Arora

Nope, while lastv is to be initialized, v is to be obtained from lastv
(which is then to be set to the new value of v) until lastv and v are
close
to each other. A small addon- it says subscripting error. "Vector
required"
was when I used v[i],i=1-5 on the LHS of my equations.

Nick Cox

Well, in this version v appears to be undefined. Perhaps you intended to
initialise v=J(1,5,0). 

Nick 
n.j.cox@durham.ac.uk 

Abhimanyu Arora

Thanks for your reply
Yes, indeed. Here is the code, when I tried to use the do-while command.
But, the problem with this is that it gives,"error 3201, vector
required"
which is surprising.
Kind regards
Abhimanyu

mata

numeric vector fpt(real scalar x,real scalar y)
{

lastv=J(1,5,0)

do{

v[1,1]=0.5775+ln(exp(x+0.9*lastv[2])+exp(y+0.9*lastv[1]))
v[1,2]=0.5775+ln(exp(2*x+0.9*lastv[3])+exp(y+0.9*lastv[1]))
v[1,3]=0.5775+ln(exp(3*x+0.9*lastv[4])+exp(y+0.9*lastv[1]))
v[1,4]=0.5775+ln(exp(4*x+0.9*lastv[5])+exp(y+0.9*lastv[1]))
v[1,5]=0.5775+ln(exp(5*x+0.9*lastv[5])+exp(y+0.9*lastv[1]))

lastv=v

}while (mreldif(v,lastv)<10^(-6))
return(v)
}


fpt(2,3)


end

Nick Cox

This doesn't look complete. 

-mreldif(,)- will surely complain if -lastv- is not defined when it is
called. 

Is there more code upstream of this? 

Conversely, my guess is that -lastv- has been defined previously, but is
of different dimensions: hence the conformability message. 

You might try some initialisation of -lastv- to 1,1,1,1,1 before
entering the loop. 

Nick 
n.j.cox@durham.ac.uk 

Abhimanyu Arora

I am trying to use mata to find the fixed point of a contraction
mapping.
Basically I have 5 non-linear equations in 5 variables (all explicitly
expressed in terms of the others). Since I wasn't successful with
'optimize', I tried to use the following idea-start with a vector of
zeros
and put the elements of vector of unknowns into the RHS of the equations
till convergence is achieved (updated value 'almost equal' to previous
one).


Following is my code and output:



clear mata

. mata
------------------------------------------------- mata (type end to
exit)
------------------------------------------------------------------------
----
--------------------------
: 
: real vector fpt(real scalar x,real scalar y) \\x and y are the
parameters
of the equations 
> {
> v=0,0,0,0,0
> 
> while (mreldif(v,lastv)<10^(-6)) {
> 
> 
> lastv=v
> 
> v[1,1]=0.5775+ln(exp(x+0.9*v[1,2])+exp(y+0.9*v[1,1]))
> v[1,2]=0.5775+ln(exp(2*x+0.9*v[1,3])+exp(y+0.9*v[1,1]))
> v[1,3]=0.5775+ln(exp(3*x+0.9*v[1,4])+exp(y+0.9*v[1,1]))
> v[1,4]=0.5775+ln(exp(4*x+0.9*v[1,5])+exp(y+0.9*v[1,1]))
> v[1,5]=0.5775+ln(exp(5*x+0.9*v[1,5])+exp(y+0.9*v[1,1]))
> 
> 
> }
> return(lastv)
> }
note: variable lastv may be used before set

: 
: 
: fpt(2,3)
               mreldif():  3200  conformability error
                   fpt():     -  function returned error
                 <istmt>:     -  function returned error
r(3200);

: 
: 
: end
------------------------------------------------------------------------
----
------------------------------------------------------------------------
----
------------------------


The error code indicates that there is some operation that is not
compatible
with the matrices involved. It would be nice if you could point out the
mistake.

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