Fix anova2 empty input handling

[sahilphad07-sudo]

anova2 failed with an undefined RIdx error when the input x was empty.

This change adds handling for the empty-input case and returns [NaN, NaN], matching MATLAB behavior.

Before

octave:> anova2([], 1, "off")
error: 'RIdx' undefined

After

octave:> anova2([], 1, "off")
ans =

   NaN   NaN

MATLAB Behavior

>> anova2([], 1, "off")
ans =

   NaN   NaN

Regression Test Added

• Empty input case

Test Results

All existing and new anova2 BISTs pass:

[pr0m1th3as]

Thank you for looking into this, but I am closing this without merging.

Returning [NaN, NaN] covers only one of the cases MATLAB handles: it gives three p-values when reps > 1, keeps the class of a single input, and returns atab and stats, so [p, t, s] = anova2 ([], 1, 'off') still crashed here. Since an empty design has nothing to analyse, anova2 now refuses it with a clear error, which closes #497.

A kind request for future contributions: please test the whole surface a change touches before opening a PR, every output and the neighbouring cases, not only the one reported. A partial fix costs more review time than it saves, and we would much rather receive fewer PRs that are complete.

[sahilphad07-sudo]

Thank you for the detailed feedback, @pr0m1th3as From next time, I’ll make sure to test the full behavior surface, including neighboring and edge cases, before opening future PRs. I appreciate the guidance.