Now whitens the input before performing PQM test

[ConnorStoneAstro]

Inputs are normalized to mean=0 and variance=1 based on the y_samples which are the reference datapoints. This is done before the reference points are taken, not sure if this matters?

[SammyS15]

I don't think it matters if it is normalized. I have tested PQMass on when images were 0 to 255, and it works fine. The important element is that both x_samples and y_samples must have the same range.

[ConnorStoneAstro]

I don't think it matters if it is normalized. I have tested PQMass on when images were 0 to 255, and it works fine. The important element is that both x_samples and y_samples must have the same range.

You are right, the normalization doesn't matter for data like that. The reason to normalize would be if there is one dimension with a range of 0 - 1000000 and the other dimensions with a range of 0 - 1. Then the first dimension will dominate all the distance calculations and the other dimensions could be very biased but that would never get detected. I talked with @LaurencePeanuts today though and we realized that in a lot of cases, the above scenario is actually fine, and you wouldn't even want to detect a bias in the tiny dimensions. So I set the whitening to False by default, but a user can turn it on if they want it.

[SammyS15]

Good after out earlier discussion