Accuracy of Quadratic Regression

[Werner]

(08-21-2023 09:33 PM)bernhardSA Wrote:  A high school student may not care [...]

A high school student should be aware of what they are doing.
Calculation devices have limits, and we run into one here.
My regular MLR/POLR regression programs for the 42S fail utterly here, not even getting a single digit correct; the version with orthogonal factorization gets 5-6 digits correct (and is thus already better than the Prime's result in HOME).
But the ultimate lesson is what Albert shows us here (and in a zillion other posts): have a look at the input data, and remove possible causes of numerical instability or inaccuracy.

Cheers, Werner