Simple econometric regression

Hi,

I would like to perform a simple econometric regression using Matlab, but I have been struggling to do it for a few days now. Quite simply, I have a set of data for both X and Y. I would like to perform a regression of the following form: Y= BetaZero + Beta*X + ErrorTerm. Could anyone please tell me what is the right function to do so? I've been trying to accomplish that by using regress(), but I realised that it does not return the value of the constant term (aka Beta zero or alpha). Thank you very much for your help.

Answer 1

[Oleg Komarov]

Use regstats then with the same inputs:

regstats(y,x,'linear')

[Marcin Kuc]

Perfect. Thanks a lot. It gives so much information and I was wondering how is it different to:

Betas = [ones(length(X),1) X] \ y

[Oleg Komarov]

It uses qr decomposition but it's pretty much the same.

[Shashank Prasanna]

If you open up regstats or regress in MATLAB using the edit command you will see that it does the same as what you describe with \ operator. An alternative and recommended approach is the new LinearModel.fit(X,y) which will give you the result and the fit statistics. Make sure you are using MATLAB R2012a and above

Answer 2

[Fangjun Jiang]

You have the right function. Just try it with three return variables.

[B,BINT,R] = REGRESS(Y,X) returns a vector R of residuals.

Read the full text of help regress

[Marcin Kuc]

Ok, as for the results that it returns, which ones will be beta0 and beta1? I believe B should be Beta1 and R is an array with residuals. Where's the constant term?

[Oleg Komarov]

You have to add a column of ones for the constant. As already suggested read the documentation.

[Marcin Kuc]

I tried to understand it, but it appears to be far too complicated for typical Eviews users. I will definitely dive into it after I finish my current project.

However, I would be very grateful if you could tell me exact way of getting the constant. Thank you in advance.

Answer 3

[Fangjun Jiang]

To use regress();

x=(1:100)';
y=10+2*x+rand(100,1);

[B,BINT,ErrorTerm]=regress(y,[ones(size(x)),x]);
BetaZero=B(1)
Beta=B(2)

BetaZero =

     10.6381

Beta =

      1.9976