jmaih
commented
4 years ago I was able to perform the following estimation using a different optimizer. fmincon is not a good enough optimizer for your problem. Consider for instance using the following results as initial conditions in your estimation problem.
` MODEL ESTIMATION RESULTS distribution initval mode mode_std
g BETA 0.65 0.91646 0.0080197
ksi GAMMA 3 3.049 0.057617
fai_k GAMMA 60 96.173 20.264
rho_am BETA 0.6 0.97439 0.10356
sig_am INV_GAMMA 5.09 1.8185 0.1138
fai_p GAMMA 60 102.36 24.083
re BETA 0.8 0.68121 0.11456
de_g BETA 0.3 0.29756 0.073699
de_d BETA 0.3 0.29638 0.074403
rho_o BETA 0.6 0.77455 0.10562
sig_o INV_GAMMA 9.81 8.3267 0.83134
rho_ao BETA 0.6 0.69619 0.097327
sig_ao INV_GAMMA 6.3 5.9955 0.26672
rho_tao BETA 0.6 0.95831 0.0567
omega_y GAMMA 0.5 0.55896 0.15041
omega_pai GAMMA 1.5 3.001 0.83377
omega_r GAMMA 0.5 0.38651 0.13968
sig_tao INV_GAMMA 10 7.9058 0.51555
____________ _______ _______ _________
distribution initval mode mode_std log-post: -354.7995 log-lik: -309.4384 log-prior: -45.3611 log-endog_prior 0.0000 numberOfActiveInequalities 0 log-MDD(Laplace) -386.0387868 estimation sample is: 1996Q1 : 2015Q4 (80 observations) solution algorithm is: rise_1 estimation algorithm is: bee_gate number of estimated parameters is: 18 number of function evaluations is: 32806 `
Dear Junior Maih: Is the result reliable with only 2 trials？Could you please tell me how to deal with the problems of trials and singularity? Here is my code： 1.zip Many thanks! Kidd June 26, 2020