verbatim; options_.noprint = true; number_of_simulations = 1; calibrated_values = M_.params; Sigma_e = M_.Sigma_e; options_.bnlms.set_dynare_seed_to_default = false; Beta = zeros(number_of_simulations, M_.param_nbr); for i=1:number_of_simulations % Set initial conditions randomly firstobs = rand(3, length(M_.endo_names)); % Set parameters to calibrated values (because after the % estimation parameters in equation 7 are updated with OLS % estimator). M_.params = calibrated_values; M_.Sigma_e = Sigma_e; % Simulate the model. simdata = simul_backward_model(dseries(firstobs, dates('1995Q1'), M_.endo_names), 1100); % Select a subsample. simdata = simdata(simdata.dates(101:1100)); % Get vector of indices to exclude exogenous variables from the dataset. names = regexp(simdata.name, 'res\w*'); idxs = find(cellfun(@isempty, names)); % Perform the estimation of equation 7. if strcmp(M_.fname, 'ols_param_names') dyn_ols(simdata{idxs}, {}, {'eq7'}, '', {{'de_g_yer_ecm_u2_stn_L1', 'de_g_yer_de_g_yer_L1'}}); elseif strcmp(M_.fname, 'ols_date_range') dyn_ols(simdata{idxs}, {}, {'eq7'}, '', {}, simdata.dates(2:end-3)); else dyn_ols(simdata{idxs}, {}, {'eq7'}); end if i ~= number_of_simulations oo_ = rmfield(oo_, 'ols'); end % Store the estimation results in Beta Beta(i, :) = M_.params'; end % Get the indices of the estimated parameters pid = oo_.ols.eq7.param_idxs; if number_of_simulations > 1 if max(abs(calibrated_values(pid)-mean(Beta(:,pid)', 2)))>1e-2 error('There is probably an error in the OLS routine.') end else % NB: estimation results depend on the sequence of generated random % numbers, hence we need to hardcode different results between MATLAB % and Octave if strcmp(M_.fname, 'ols_base') if isoctave good = [2.509401520579608e-02 -1.212468793686628e-01 2.983003417202038e-02 4.645373135672548e-01 -6.312448906643650e-02]; else good = [0.025487074270635 -0.126705792250240 0.002639663639062 0.486913131141265 -0.027702100375691]; end elseif strcmp(M_.fname, 'ols_param_names') if isoctave good = [-0.125307512725951 0.465364467369817]; else good = [-0.125265361034515 0.497003481797959]; end elseif strcmp(M_.fname, 'ols_date_range') if isoctave good = [4.775810440061960e-01 2.096533656963026e-02 3.264889547460539e-02 -1.167420901882959e-01]; else good = [0.468607715487806 -0.011224442004140 0.033681224841824 -0.126743281213504]; end elseif strcmp(M_.fname, 'ols_wc_1') if isoctave good = [4.778519485900916e-01 2.028231364527827e-02 3.186766890724485e-02 -1.173897445629599e-01]; else good = [0.466534186307255 -0.010596174928406 0.035098464382347 -0.126502172176703]; end elseif strcmp(M_.fname, 'ols_wc_2') if isoctave good = [4.679679703340907e-01 2.987618894535485e-02 2.503849196608867e-02 -1.206391201296474e-01 3.435379133407408e-02]; else good = [0.485802603276717 0.002660854661647 0.025434782170432 -0.126939890810119 0.043688440124315]; end elseif strcmp(M_.fname, 'ols_wc_3') if isoctave good = [4.766230170524117e-01 2.500479091401997e-02 2.333265887171398e-02 -9.913848386063488e-02]; else good = [0.490636551206316 0.014209413280418 0.028534044956486 -0.094387460262917]; end end if max(abs(good-Beta(:,pid)'))>1e-2 error('There is probably an error in the OLS routine.') end return end for i=1:length(pid) figure(i) hold on title(strrep(M_.param_names(pid(i),:), '_', '\_')); bandwidth = mh_optimal_bandwidth(Beta(:,pid(i)), length(Beta(:,pid(i))), -1, 'gaussian'); [abscissa, f] = kernel_density_estimate(Beta(:,pid(i)), 256, length(Beta(:,pid(i))), bandwidth, 'gaussian'); plot(abscissa, f, '-k', 'linewidth', 2); line([calibrated_values(pid(i)) calibrated_values(pid(i))], [0 max(f)*1.05], 'LineWidth', 2, 'Color', 'r'); line([mean(Beta(:,pid(i))) mean(Beta(:,pid(i)))], [0 max(f)*1.05], 'LineWidth', 2, 'Color', 'g'); axis tight box on hold off end end;