jst a thot....what if the standardisd deviation is zero...shd it be den consisered positive,negative or jst ignored in calculatin probabilistic values...
A few thoughts here: - Signs tests (and grouping of signs tests for that matter) where there are zero deviations are not covered in Core Reading. I'd be surprised to see this examined, and don't think it ever has been. - Since there's nothing explicitly in the course about it, the important thing would be to do something sensible based on the situation. This should gain credit. So if, for instance, we had 10 positive deviations, 1 zero and 5 negatives, then I'd probably calculate the p-value of the test as 2xP(P>=10), where we multiply by 2 as it's a two-sided test and the number of positive deviations has distribution Bin(16, 0.5). - If it came up on an exam and I had a choice of which tests to perform, I'd steer clear of those depending on sign. - Remember that a zero deviation can be a good thing in a graduation. It means that the graduated rates fit very well to the underlying data at that point. Lots of zero deviations may indicate a lack of smoothness, but excellent adherence to data.
