Error in RSS calculus in LinReg function

[Christian Paz]

When we run the following code, it is expected that the RSS value given by the Estimate and the LinReg function take the same value. Instead, LinReg returns 11.4009263094 and Estimate returns (the correct value) 0.2280185256.

This error may lead to incorrect decisions based on RSS measure. We think that it is calculated as the square of sigma, it is (n-p) * sigma (n: number of data, p: number of parameters).

Simulated data y = alfa+beta X+e

Using LinReg RSS (residual square sum ) is grown!

Real n=50; lenght of the simulated sample
Real alfa=1;
Real beta=2;
Real Sigma=0.5;
Matrix X1 = SetCol(NCopy(n, 1));
Matrix X2 = Rand(n,1,0,1);
Matrix X =X1|X2; explanatory Matrix
Matrix Y = RProd(X1, alfa)+RProd(X2, beta)

+ Gaussian(n, 1, 0, Sigma); response Matrix

Set Resp=LinReg(Y,X);

Using Estimate is fine!

Serie Yr = SetSer(MatSet(Y),y2005m02,Mensual);
Serie Const = CalInd(C,Mensual);
Serie Xr = SetSer(MatSet(X2),y2005m02,Mensual);

Set Input_1 = SetOfSet(
InputDef(0.1, Const),
InputDef(0.1, Xr)
);

Set SetAR = SetOfReal(1,1); Parte AR de ModelDef
Set SetMA = SetOfReal(1,1); Parte MA de ModelDef
Polyn Diferen = 1;

Set Model1 = ModelDef(

Yr, Serie Output,
1, Real FstTransfor,
0, Real SndTransfor,
12, Real Period,
0, Real Constant,
Diferen, Polyn Dif,
SetAR, Set AR,
SetMA, Set MA,
Input_1, Set Input,
Copy(Empty) Set NonLinInput

);

Set Resp2 = Estimate(Model1);

[Víctor de Buen Remiro]

We are working about this problem

Thanks to report it

[Víctor de Buen Remiro]

Sets AR and MA must contain polynomials instead of real numbers.

Set SetAR = SetOfPolyn(1,1); Parte AR de ModelDef
Set SetMA = SetOfPolyn(1,1); Parte MA de ModelDef

I will try to see way Estimate doesn't shows the corresponding error message.
Also, I will try to unify criteria for diferent informative statistics as RSS.

For example, RSS of LinReg is returning RSA (Residuals Sum Average)

[Víctor de Buen Remiro]

Problem has been solved in CVS HEAD (1.1.7)