http://shahramabyari.com/wp-content/uploads/2015/12/codedDiamondss.csv

]]>In the example, you have used codedDiamond.csv. Can you please provide the codedDiamond.csv as I have downloaded the diamond.csv from ggplot which you code does not work up. I will appreciate if you can provide codedDiamond.csv.

Thank you ]]>

sorry for delay… It has been a busy week. You are right… The paper recommends using probability for t-test and supremum for KS test. However, the general idea is finding the probability that the distributions are the same and the differences are because of randomness and one of the strengths of method proposed in paper is that you can use alternative methods for quality criterion. I think 1-p should be a good proxy to check if distributions are the same.

Thanks,

~Shahram

thanks for the code, however I do not understand why you use the asymptotic p value. The paper says the supremum of the difference in cumulative marginal and conditional distribution. I am implementing in matlab so I use the test statistic k and not the p.

Can you please explain why you use p and subtract it from 1??

Thanks

Arvind