Closed Serabe closed 12 years ago
And increasing the size of the sample to 50000 or 100000?
This has nothing to do with false negatives. Kolmogorov-Smirnov is not valid for discrete distributions so another solution needs to be found.
I am using the goodfit function in vcd package (take a look at testsuite.R), but for getting the p value I need to call summary, which produces output as a side effect.
+1 to goodness of fit for discrete distributions
Current test produces non-desired outputs and lost warnings.
I don't like it but I cannot find any other way.