How to get regression statistics for several run with different data sets?

I have a data set of 48 row and 5 column. Like the following Year Month Water_Demand Population rainfall, data for 4 years (48 months). Here water_demand is dependent variable and population and rainfall are independent variables. I estimate the regression coefficients four times, each time excluded 12 months from the data set.Here is the code, number_of_lines=length(data); C = zeros(3,4); for n=[1 13 25 37] under_test=data; under_test(n:n+11,:)=[]; B1=under_test(:,3); B2=under_test(:,4); B3=under_test(:,5); x1=[ones(size(B1)),B2,B3]; y1=B1; C(:,n)=regress(y1,x1); end now i want to get standard error, R2, adjusted R2 of those four run. Can any one help plz.thanks.

Answer 1

[Richard Willey]

Hey there

The following code assumes that you have the 12a release of Statistics Toolbox. (In other words, I am using the new LinearModel function which makes life much much easier)

I'm going to generate some "fake" data. It should be pretty easy to modify this for your own uses.

% Generate a data set

X1 = randn(100,1);
X2 = randn(100,1);
X3 = randn(100,1);
Y = 3 + 5*X1 + 7*X2 + 9*X3 + randn(100,1);

X = [X1 X2 X3];

% Use Linear Model to model Y as a function of X
myFit = LinearModel.fit(X, Y);

% Display "myFit"
% Note that R^2, Adjusted R^2, and the standard error of the regression
% coefficients are all part of the default display
disp(myFit)

% See all the methods available for myFit
methods(myFit)

% Retrieve the statistics of interest

R_Squared = myFit.Rsquared.Ordinary
Adj_R_Squared = myFit.Rsquared.Adjusted
SE = myFit.Coefficients.SE

[MD Mahmudul Haque]

Hi Richard, thanks for your answer. I can get the required statistics for an individual run by your code. But if you see my question, i am running inside the loop. Loop has been defined to run required times by excluding the 12 months data each time. In my case, loop run 4 times, as the data length is only 48 row. I get the matrix of coefficient of four run at a time. My question is, is it possible to get the other statistics in that matrix or in a separate matrix for those four runs together.

[Tom Lane]

In your example you assigned coefficients into C(:,n) each time through the loop. You could do the same with SE, R^2, etc. by assigning into arrays of those. Richard showed how to use the new LinearModel feature to get those results. You could also use regstats to get them.