www.gusucode.com > stats 源码程序 matlab案例代码 > stats/LinearModelUsingStepwiseRegressionExample.m

    %% Fit a Linear Model Using Stepwise Regression

%% 
% Load the sample data.
load hald

%%
% |hald| contains hardening data for 13 different concrete compositions.
% |heat| is the heat of hardening after 180 days. |ingredients| is the percentage
% of each different ingredient in the cement sample.

%% 
% Fit a linear model to the data. Set the criterion value to enter the model
% as 0.06.
 mdl = stepwiselm(ingredients,heat,'PEnter',0.06)

%%
% By default, the starting model is the constant model. |stepwiselm| performs
% forward selection and |x4|, |x1|, and |x2|, respectively, as the corresponding
% $p$-values are less than the |PEnter| value of 0.06. |stepwiselm| later
% uses backward elimination and eliminates |x4| from the model because,
% once |x2| is in the model, the $p$-value of |x4| is higher than
% the default value of |PRemove|, 0.1.