Octave / Matlab implementation of confidence intervals in multiple regression

Question

I need to implement confidence intervals for multiple regression coefficients in Octave / Matlab.

The problem is defined in a general way: data Y, projection matrix X, coefficients β, so Y = βX. Then the code for β is simple:

beta = pinv(X)*Y

Now, being a stupid physicist, I lost a little confidence and intervals of prediction. Formulas, as well as their implementation.

Note: I know there is a Matlab mvregress function, but it is still missing from Octave, which I actually use.

Note 2: This question was asked at CrossValidated and is marked as a disabled section that focuses on programming.

+5

Victor pira Oct 15 '17 at 7:40

1 answer

Andre · Answer 1 · 2017-10-23T05:10:06+0000

I think this is what you want to find:

 [b, bint, r, rint, stats] = regress (y, X, [alpha]).

where bint is the confidence interval for beta.