I am working on some robust statistics for moran coefficients with @weikang9009 and I've discovered the two sided null is degenerate when the entire reference distribution is exactly the same.
In this case, the two sided null using percentiles will be degenerate here, since percentiles are degenerate.
So, I think we should ditch the percentile and use the equivalent (and more robust) 2*(min(greater, lesser) +1)/(permutations+1)
I didn't consider this degenerate case. Sorry!
I am working on some robust statistics for moran coefficients with @weikang9009 and I've discovered the two sided null is degenerate when the entire reference distribution is exactly the same.
In this case, the two sided null using percentiles will be degenerate here, since percentiles are degenerate.
So, I think we should ditch the percentile and use the equivalent (and more robust) 2*(min(greater, lesser) +1)/(permutations+1)
I didn't consider this degenerate case. Sorry!