T test p values for regression coefficients

Hi. I am trying to fit a linear model Y= m*X. I wanted to get T test p values for individual regression coefficients. I have seen that the function regstat does provide the T test p values. The problem is that while performing regression , regstat adds a column of ones by itself to the feature set (X). I do not plan to include the column of ones as my model is simple Y=m*X instead of Y=m*X + c. Is there any way or any function I could use to compute the T test p values without including a column of ones in the feature set.

Thanks

Answer 1

[Shahid]

Hi. So I found a solution. There is this library posted by someone http://www.mathworks.com/matlabcentral/fileexchange/26169-regstats2

It is an improved version of original regstats function in matlab and also allows p values of coefficient for regression with intercept term.

Answer 2

[Tom Lane]

If you look at "help regstats" you will see that the default model is 'linear' and this includes the constant. But you will also see that there's a way to specify the terms you want directly. An identity matrix will provide what you need. Admittedly this is obscure, but try this:

>> x = rand(20,3);
>> y = x*[1;-2;3] + randn(20,1)/100;
>> s = regstats(y,x,'linear'); s.beta'
ans =
    0.0117    0.9907   -2.0019    2.9866
>> s = regstats(y,x,eye(3)); s.beta'
ans =
    0.9982   -1.9984    2.9977

I show the coefficient estimates, but the t statistics and p-values are in there as well.

If you have a recent release of MATLAB, consider

LinearModel.fit(x,y,'Intercept',false)

Answer 3

[the cyclist]

I believe you could use the function nlinfit() from the Statistics Toolbox to do this.

Here is an example of the function:

rng(1)

% Here is an example of using nlinfit(). For simplicity, none of
% of the fitted parameters are actually nonlinear!

% Define the data to be fit
x=(0:1:10)'; % Explanatory variable
y = 5 + 3*x + 7*x.^2; % Response variable (if response were perfect)
y = y + 2*randn((size(x)));% Add some noise to response variable

% Define function that will be used to fit data
% (F is a vector of fitting parameters)
f = @(F,x) F(1) + F(2).*x + F(3).*x.^2;
F_fitted = nlinfit(x,y,f,[1 1 1]);

% Display fitted coefficients
disp(['F = ',num2str(F_fitted)])

% Plot the data and fit
figure
plot(x,y,'*',x,f(F_fitted,x),'g');
legend('data','fit')

My example only uses one explanatory variable, and does have an intercept term, but you should be able to adapt it to your model.

Answer 4

[Shahid]

Hi. Thanks for the answer. I actually am interested in getting the t-test P values of the coefficients as returned by REGSTATS function in form of stats.tstat

[the cyclist]

For your future reference, this "answer" would have been better placed as a comment on my answer.