DM-18638: Check for at least 2 visits in AMx calc

[yalsayyad]

before including a pair's distance stdev. The distance of a pair that
appears in only one visit has a std of zero, biasing the median RMS
if used.

[RobertLuptonTheGood]

The reason for using the unbiased stdev 1/(n-1) \sum(x - \bar x)^2 is partially to avoid this problem. Are you using 1/n?

[yalsayyad]

We need at least 2 values regardless of what estimator we use.

Looking at the line below, it uses np.std, which I know you've complained uses 1/n by default:
https://docs.scipy.org/doc/numpy/reference/generated/numpy.std.html

[wmwv]

Yes, this was my fault. I thought that numpy.std used 1/(n-1) and would return NaN for n=1. But I did not test this implicit assumption.

I suggest we require 2 values and then return n/(n-1) * np.std(...) to correct the bias.

[yalsayyad]

@wmwv Do want to file a ticket and push that change or shall I?