www.gusucode.com > econ 案例源码程序 matlab代码 > econ/AssessModelStabilityWithSmallComplementarySubsampleSizeExample.m
%% Assess Regression Model Stability With Small Complementary Subsample Size % The break point Chow test is not appropriate for instances when the % the complementary subsample has as many or fewer observations as % coefficients in the model. For these instances, use the forecast % test. %% % Load the U.S. food consumption data set. Extract the food price index % (|P|) and food consumption index (|Q|). Suppose that data for 1927 % through 1949 are available. % Copyright 2015 The MathWorks, Inc. load Data_Consumption dates = (1927:1949)'; Tbl = DataTable(cellstr(num2str(dates)),{'P','Q'}); figure; plot(dates,Tbl{:,{'P' 'Q'}},'o-') axis tight grid on xlabel('Year') ylabel('Index') title('{\bf Time Series Plot of All Series}') legend({'Price','Consumption'},'Location','SE') %% % Measurements are missing from 1942 through 1947, which correspond to % World War II. A common assumption is that consumption elasticity is a linear % function of price elasticity (and perhaps other variables). %% % Identify the indices before World War II. preWarIdx = (dates <= 1941); %% % Because each measurement is already normalized, obtain elasticities by % applying the log transformation. Plot consumption elasticity as a % function of price elasticity. Tbl.LQ = log(Tbl.Q); % Consumption elasticity Tbl.LP = log(Tbl.P); % Price elasticity figure; plot(Tbl.LP(preWarIdx),Tbl.LQ(preWarIdx),'bo',... Tbl.LP(~preWarIdx),Tbl.LQ(~preWarIdx),'r*'); axis tight grid on lsline; xlabel('Price elasticity') ylabel('Consumption elasticity') legend('Pre-war observations','Post-war observations',... 'Location','best') %% % Because the slopes and intercepts between the subsamples appear to be % different, the plot suggests to reject model stability. %% % Conduct a forecast test including all coefficients, and % separately for each coefficient. bp = find(preWarIdx,1,'last'); Coeffs = [1 1; 1 0; 0 1]; chowtest(Tbl(:,{'LP','LQ'}),bp,'Test','forecast','Coeffs',Coeffs); %% % All tests fail to reject the null hypothesis of model stability.