kalman normailized innovation squared distribution is unexpected #1517
Description
The filter notes "Chi-squared threshold for 3 degrees of freedom, p=0.01", but I think the normalized innovation squared (NIS) should have 2 degrees of freedom as we are measuring position on a surface. The mean of the chi-square distribution should be the number of degrees of freedom.
Furthermore, I can't find a data set where the distribution of the NIS values is anywhere close to a chi-square distribution. The values traced are many orders of magnitude too small.
This is a mystery.