Re: MIMO linear regression

[CdeMills]

Ismael Diego Nunez-Riboni wrote
>> I would like to probe further and test
>> wether or not the model coefficients are significant. 
> 
> I'm by far no expert in signal analysis or statistics and probably I'm
> also overlooking something or misunderstanding your problem, but if you
> are making such linear models with Octave and care about the significance
> of your regression parameters, why not to use the function "regress"? : 
> 
> [B, BINT, R, RINT, STATS] = regress (Y, X, [ALPHA])
> 
> BINT is the confidence interval of your regression coefficients calculated
> with the confidence level ALPHA (ALPHA = 0.05 means the 95% confidence
> level, if I'm right).

Hello Ismael,

thanks for the suggestion. Unfortunatelly, like many examples I've found,
the regression is between a single variable and many explanatory variables,
i.e. Y must be a vector. OTOH the extension to matrix Y should be
straightforward.

Regards

Pascal



--
View this message in context: 
http://octave.1599824.n4.nabble.com/MIMO-linear-regression-tp4648728p4648754.html
Sent from the Octave - General mailing list archive at Nabble.com.