time-varying coefficients in a linear model
Posted: Tue Apr 23, 2013 6:44 am
I want to estimate a linear model with time-varying coefficients
Y_t = alfa_t + beta_t*X1_t + lambda_t*X2_t
The programa I use is the following
*THE RESULT IS THE FOLLOWING
*AS CAN BE SEEN THE ESTIMATION DOES NOT PROGRESS FROM ITS INITIAL VALUES.
*HOW CAN I IMPROVE THE PROGRAM?
I am sure it is a basic mistake. I write in this forum after several days trying to find a solution by myself. Any suggestion is very welcomed.
Y_t = alfa_t + beta_t*X1_t + lambda_t*X2_t
The programa I use is the following
Code: Select all
Allocate 574
open data c:\HED9809.xls
data(format=xls,org=obs) / Y X1 X2
nonlin alfa beta lambda v walfa wbeta wlambda
*I introduce the following values for initialization
compute ALFA = -0.001, BETA = 0.6, LAMBDA = 1.0, V = 0.002,WALFA = 0.0004, WBETA = 0.8, WLAMBDA=0.35
declare symmetric sw
compute sw=%DIAG(||walfa,wbeta,wlambda||)
dec frml[rect] a
frml a = ||1.0, 0.0, 0.0|0.0, 1.0, 0.0|0.0, 0.0, 1.0||
dlm(y=lnds,c=||1.0,X1,X2||,a=a,sv=v,sw=sw,type=smooth,method=BFGS,condition=4) / xstates vstates*THE RESULT IS THE FOLLOWING
Code: Select all
DLM - Estimation by BFGS
Convergence in 7 Iterations. Final criterion was 0.0000027 <= 0.0000100
Usable Observations 570
Rank of Observables 570
Log Likelihood 620.6579
Variable Coeff Std Error T-Stat Signif
************************************************************************************
1. ALFA -0.001000000 0.000000000 0.00000 0.00000000
2. BETA 0.600000000 0.000000000 0.00000 0.00000000
3. LAMBDA 1.000000000 0.000000000 0.00000 0.00000000
4. V 0.000579180 0.000122125 4.74253 0.00000211
5. WALFA 0.000400000 0.000000000 0.00000 0.00000000
6. WBETA 0.800000000 0.000000000 0.00000 0.00000000
7. WLAMBDA 0.350000000 0.000000000 0.00000 0.00000000*AS CAN BE SEEN THE ESTIMATION DOES NOT PROGRESS FROM ITS INITIAL VALUES.
*HOW CAN I IMPROVE THE PROGRAM?
I am sure it is a basic mistake. I write in this forum after several days trying to find a solution by myself. Any suggestion is very welcomed.
