[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

From |
"Randy Akee " <akee@iza.org> |

To |
<statalist@hsphsun2.harvard.edu> |

Subject |
RE: st: Mata and Optimize Command |

Date |
Mon, 24 Sep 2007 18:04:06 +0200 |

Thanks for this, this was extremely helpful. R -----Original Message----- From: owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Jeff Pitblado, StataCorp LP Sent: Montag, 24. September 2007 16:49 To: statalist@hsphsun2.harvard.edu Subject: Re: st: Mata and Optimize Command Randy Akee <akee@iza.org> is getting a conformability error in the -md0()- Mata routine he wrote for use with -optimize()-: > I am trying to use the MATA Optimize command to get a minimum distance > estimator of some parameters of interest. I have six equations and six > unknown parameters - my actual problem is a bit more complex, but I > can't seem to get the simple version to work below: > > I've typed in two matrices, c and f. I'm trying to find the parameters, > p, that minimize: (c-f(p))'(c-f(p)). > > The error that I get is that the function is not conformable, does > anyone have a suggestion on how I could correct this in what I've done? > I believe the issue is that I need somehow to specify the dimensions of > the parameter vector, p. I'm just not sure if that is possible to do, > any ideas would be appreciated. Randy also reported the Mata code that reproduces the error. Here is Randy's evaluator function: > void md0(todo, p, c, f, lnf, S, H) > { > lnf = (c-f*p)'*(c-f*p) > } Here is a brief description of the arguments to this function: Input variables: ---------------- todo -- a scalar message variable from -optimize()- that indicates whether to compute 1st and 2nd order derivatives; this variable can safely be ignored if your routine is not going to compute derivatives p -- the current value of the parameter vector (a rowvector) c -- Randy's first user-defined argument > : c > 1 > +-----+ > 1 | 1 | > 2 | 2 | > 3 | 3 | > 4 | 4 | > 5 | 5 | > 6 | 6 | > +-----+ f -- Randy's second user-defined argument > : f > 1 2 3 4 5 6 > +-------------------------+ > 1 | 1 1 0 0 0 0 | > 2 | 0 1 1 0 0 0 | > 3 | 0 0 1 1 0 0 | > 4 | 0 0 0 1 1 0 | > 5 | 0 0 0 0 1 1 | > 6 | 1 0 0 0 0 1 | > +-------------------------+ Output variables: ----------------- lnf -- the value of the objective function that is being optimized g -- the gradient vector H -- the Hessian matrix The problem with Randy's evaluator it that with 'c' a 6x1 column vector and 'f' a 6x6 matrix, 'p' would need to be a 6x1 column vector. There are two reasons why this is a problem: 1. -optimize()- requires that 'p' is a rowvector 2. Randy set the starting values using > : optimize_init_params(S,(0,0)) which is a 1x2 rowvector. In this case, 'p' needs to be a 1x6 rowvector, so there are missing transpose operators in Randy's -md0()- function. Here is how I would code Randy's evaluator and starting values: void md0(todo, p, c, f, lnf, S, H) { real colvector diff diff = c - f*p' lnf = cross(diff,diff) } optimize_init_params(S, J(1,6,0)) [-cross(z,z)- is faster that -z'*z-] Note that 'lnf' will be a scalar, but Randy coded 'md0()' as a type -v0- evaluator. If Randy wants to return a column vector of the squared differences, he can use the following function evaluator void mv0(todo, p, c, f, lnf, S, H) { lnf = (c - f*p') :* (c - f*p') } Here is the do-file I composed while looking into Randy's code. ***** BEGIN: mata: c = (1,2,3,4,5,6)' f = (1,1,0,0,0,0\0,1,1,0,0,0\0,0,1,1,0,0\0,0,0,1,1,0\0,0,0,0,1,1\1,0,0,0,0,1 ) void md0(todo, p, c, f, lnf, S, H) { real colvector diff diff = c - f*p' lnf = cross(diff,diff) } Sd = optimize_init() optimize_init_evaluator(Sd, &md0()) optimize_init_evaluatortype(Sd, "d0") optimize_init_params(Sd, J(1,6,0)) optimize_init_which(Sd, "min") optimize_init_argument(Sd, 1, c) optimize_init_argument(Sd, 2, f) optimize(Sd) void mv0(todo, p, c, f, lnf, S, H) { lnf = (c - f*p') :* (c - f*p') } Sv = optimize_init() optimize_init_evaluator(Sv, &md0()) optimize_init_evaluatortype(Sv, "v0") optimize_init_params(Sv, J(1,6,0)) optimize_init_which(Sv, "min") optimize_init_argument(Sv, 1, c) optimize_init_argument(Sv, 2, f) optimize(Sv) end ***** END: --Jeff jpitblado@stata.com * * For searches and help try: * http://www.stata.com/support/faqs/res/findit.html * http://www.stata.com/support/statalist/faq * http://www.ats.ucla.edu/stat/stata/ * * For searches and help try: * http://www.stata.com/support/faqs/res/findit.html * http://www.stata.com/support/statalist/faq * http://www.ats.ucla.edu/stat/stata/

**References**:**Re: st: Mata and Optimize Command***From:*jpitblado@stata.com (Jeff Pitblado, StataCorp LP)

- Prev by Date:
**RE: st: Goodness of fit and model comparison in xtgee** - Next by Date:
**st: Drawing 3 normally distributed variable with a given correlation structure to an existing variable** - Previous by thread:
**Re: st: Mata and Optimize Command** - Next by thread:
**RE: st: Mata and Parameter Constraints** - Index(es):

© Copyright 1996–2016 StataCorp LP | Terms of use | Privacy | Contact us | What's new | Site index |