CLS in LS equation

[Katjes]

Dear all,
we used version 9.5 and version 8.1 to perform e.g.

Code: Select all

equation  eq_1.ls(optmethod=legacy, arma=cls) hours c ma(1) ma(2) ma(3) ar(1) ar(2) @trend

that gives us a least-square estimation (without the sigma-parameter).
Both version provide the same results,

Now one of my colleagues has version 9 and the same code produces a ML estimation. We find out that the code has to be

Code: Select all

equation  eq_1.ls(optmethod=legacy, arma=eviews) hours c ma(1) ma(2) ma(3) ar(1) ar(2) @trend

to get the same results as in EViews 9.5.

But if I adopt the latter code in EViews 9.5 the code provides an ML-estimate. Very confusing!

What is the best method, that all EViews 9 versions produces least square estimates without sigma?