// --+ options: json=compute, stochastic +-- var y x z; varexo ey ex ez; parameters theta gamma; gamma = .3; model; [name='eqx'] x = x(-1)^.8*exp(ex) ; [name='eqz'] z = exp(ez); [name='eqy'] y = x^(1-gamma)*z^gamma + ey; end; histval; x(0) = 0.1; z(0) = 0.5; end; shocks; var ex = .05; var ez = .05; var ey = .01; end; // Simulate a sample. simulations = simul_backward_model([], 5000); // Set residuals in the estimated equation to NaNs. simulations.ey = dseries(NaN); // Set initial guess for the estimated parameter. clear('eparams') eparams.gamma = 0.6; // Call estimation routine. if ~isoctave % Under Octave, estimate.nls (provided by matlab/+estimate/nls.m) is not % accessible because there is a function which has the same name as the % +estimate package, namely matlab/cli/estimate.m. This is a known Octave % bug (https://savannah.gnu.org/bugs/?func=detailitem&item_id=46889). estimate.nls('eqy', eparams, simulations, 1001Y:5000Y, 'annealing') end