error when estimating with a bad defined AR or MA set of Polyn

[Jorge]

The following code will break down TOL!!!!

Set SimulateARIMA(TimeSet tms, Date from, Real size, Set input,

Real sig2, Real deg_dif, Polyn pol_ar, Polyn pol_ma)

{

Date to = Succ(from, tms, size-1);

Serie at = SubSer(Gaussian(0,sig2,tms), Succ(from,tms,-100), to);
Polyn pol_dif = If(deg_dif==0, 1, 1-B^{deg_dif);
Serie Nt = SubSer(DifEq(pol_ma/(pol_dif*pol_ar),at), from, to);
Set input_fil = EvalSet(input, Serie (Set inp) {Serie inp[1]:inp[2]});
Serie filter = BinGroup("+", input_fil);}

Set ModelDef(filter+Nt, 1, 0, deg_dif, 0, pol_dif,

pol_ar, 1?,
1, pol_ma?,
input,
AllLinear)

};

Serie GenStep(TimeSet tms, Date from, Real size, Real point)
{

Date to = Succ(from,tms,size-1);
Serie SubSer(Step(Succ(from,tms,point-1),tms), from, to)

};

Set test_one(Real a)
{

Set inputs = [[InputDef(1.2, GenStep(Mensual, y2004, 600, 150)),

InputDef(-1.3, GenStep(Mensual, y2004, 600, 300))]];

Set mod = SimulateARIMA(Mensual, y2004, 600, inputs, 1.0, 0, 1-0.7*B,1);

Set Estimate(mod, First(mod[1]), Last(mod[1]))

};

Set est = test_one(1);

[Jorge]

will be considered LATER

[Jorge]

reopen for 1.1.4

[Jorge]

now tol does not crash.

the bug was related to missing data at the beginning of the estimation

you can replace the code of test_one for the one below in order to compute the
correct initial date of estimation, see bug 81 for a related solution.

Set test_one(Real a)
{

Set inputs = [[InputDef(1.2, GenStep(Mensual, y2004, 600, 150)),

InputDef(-1.3, GenStep(Mensual, y2004, 600, 300))]];

Set mod = SimulateARIMA(Mensual, y2004, 600, inputs, 1.0, 0, 1-0.7*B,1);

Date BeginOfEstim =

{

Real ARDeg = Degree(SetProd(mod->AR));
Real OmegaDeg = SetMax(EvalSet(Traspose(Extract(mod->Input,1))[1],Degree));
Succ(First(mod->Output),Dating(mod->Output),ARDeg+OmegaDeg)

};

Set Estimate(mod, BeginOfEstim, Last(mod[1]))

};