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

From |
"Nick Cox" <n.j.cox@durham.ac.uk> |

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

Subject |
st: RE: RE: RE: Including a matrix in a monte carlo simulation |

Date |
Wed, 16 Oct 2002 21:08:55 +0100 |

Nevo, Dorit > > I'm not sure I'm going in the right direction so I'm > pasting the full > program below. I am trying to calculate the mean and > standard error of > sigma. I need quite a few calculations along the was so I tried to > generalize on the basic monte carlo example that I found in > the reference > book and online. > > By the time I create the matrix Fc should already be > calculated. It works > fine without the matrix command so I'm not sure why the > problem occurred. > > Dorit > > > program define aes_monte > version 7.0 > if "`1'" == "?" { > global S_1" mean var" > exit > } > drop _all > set obs 3000 > gen B0 = -.0353521 > gen R = .5022559 + .0403609*invnorm(uniform()) > gen Dc = -.3476753 + .0700479*invnorm(uniform()) > gen Dk = .3394081 + .0281788*invnorm(uniform()) > gen Bcc = -.0106232 + .0030351*invnorm(uniform()) > gen Bkk = .0782757 + .0023668*invnorm(uniform()) > gen Bll = .0859557 + .0030879*invnorm(uniform()) > gen Bck = .0217035 + .002353*invnorm(uniform()) > gen Bcl = .0066442 + .0040709*invnorm(uniform()) > gen Bkl = -.1821175 + .0040324*invnorm(uniform()) > gen C = 14.86249 > gen lnC = 2.69884 > gen K = 576.6429 > gen lnK = 6.357223 > gen L = 486.6226 > gen lnL = 6.187489 > gen V =exp(B0 - (1/R)*ln(Dc*(C^(-R)) + Dk*(K^(-R)) + > (1-Dc-Dk)*(L^(-R))) + Bcc*(lnC^2) + Bkk*(lnK^2) + > Bll*(lnL^2) + Bck*lnC*lnK > + Bcl*lnC*lnL + Bkl*lnK*lnL) > > gen Z = Dc*C^(-R)+Dk*K^(-R)+(1-Dc-Dk)*L^(-R) > gen Fc = V*(Dc*C^(-R-1)/Z+Bck/C*lnK+Bcl/C*lnL+2*Bcc/C*lnC) > gen Fk = V*(Dk*K^(-R-1)/Z+Bck/K*lnC+Bkl/K*lnL+2*Bkk/K*lnK) > gen Fl = > V*((1-Dc-Dk)*L^(-R-1)/Z+Bcl/L*lnC+Bkl/L*lnK+2*Bll/L*lnL) > gen Fcc = Fc^2/V - Fc/C + > V*((-R*Dc*C^(-R-2))/Z+(Dc^2*C^(-2*R-2)*R)/Z^2+2*Bcc/C^2) > gen Fkk = Fk^2/V - Fk/K + > V*((-R*Dk*K^(-R-2))/Z+(Dk^2*K^(-2*R-2)*R)/Z^2+2*Bkk/K^2) > gen Fll = Fl^2/V - Fl/L + > V*((-R*(1-Dc-Dk)*L^(-R-2))/Z+((1-Dc-Dk)^2*L^(-2*R-2)*R)/Z^2+ > 2*Bll/L^2) > gen Fck = (Fc*Fk)/V + > V*((R*Dc*Dk*C^(-R-1)*K^(-R-1)/Z^2)+Bck/(C*K)) > gen Fcl = (Fc*Fl)/V + > V*((R*Dc*(1-Dc-Dk)*C^(-R-1)*L^(-R-1)/Z^2)+Bcl/(C*L)) > gen Fkl = (Fk*Fl)/V + > V*((R*Dk*(1-Dc-Dk)*K^(-R-1)*L^(-R-1)/Z^2)+Bkl/(K*L)) > > matrix H = (0,Fc,Fk, Fl\Fc, Fcc, > Fck,Fcl\Fk,Fck,Fkk,Fkl\Fl,Fcl,Fkl,Fll) > gen detH = det(H) > matrix Hck = (0,Fc,Fl\Fk,Fkc,Fkl\Fl,Flc,Fll) > gen detHck = det(Hck) > > gen sigma_ck = ((C*Fc+K*Fk+L*Fl)/(C*K))*(-detHck/detH) > summarize sigma_ck > post `1' (r(mean)) (r(Var)) > end The bottom line here is that det(H) is a matrix function yielding one number. Putting the result in a variable with 3000 obs will just give you 3000 copies of the same number. Therefore, as I understand it, you may need, within this file, to loop over observations, to extract 16 numbers from each observation and get the determinant in each case. However, there is a yet wider context -- this is being called by something else, presumably -simul-. Despite the fearsome singularity(*) of these calculations, it looks as if everything is done row-wise: that is, should all these done as scalar operations and the 3000 replications conducted via -simul-? Nick n.j.cox@durham.ac.uk (*) pun * * 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**:**st: RE: RE: Including a matrix in a monte carlo simulation***From:*"Nevo, Dorit" <dorit.nevo@commerce.ubc.ca>

- Prev by Date:
**st: RE: RE: Including a matrix in a monte carlo simulation** - Next by Date:
**st: Select dates** - Previous by thread:
**st: RE: RE: Including a matrix in a monte carlo simulation** - Next by thread:
**st: Select dates** - Index(es):

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