kolmogorov smirnov in r - cannot compute correct p-values with ties


by joanne34567
Tags: kolmogorov smirnov, p-values, ties
joanne34567
joanne34567 is offline
#1
Jan18-13, 10:13 AM
P: 13
Hi,
I'm trying to use the kolmogorov smirnov test in in R to compare one distribution with another. I'm getting the following error: cannot compute correct p-values with ties
I think this is because the dataset that I am using has in the first instance around 2000 values, and the second, 50000 values. A number of these are inherently going to be replicates of another number in the series (i.e. "ties"). I'm just wondering if this has implcations for the p value I have? Is the p value valid?
Cheers all
Phys.Org News Partner Science news on Phys.org
Better thermal-imaging lens from waste sulfur
Hackathon team's GoogolPlex gives Siri extra powers
Bright points in Sun's atmosphere mark patterns deep in its interior
Stephen Tashi
Stephen Tashi is offline
#2
Jan19-13, 11:34 AM
Sci Advisor
P: 3,173
Did you read the R documentation for the ks.test function?

You could test the sensitivity of the p-value empirically. Add small random perturbations to your data to generate new data sets that have more digits. See how much the p value changes.


Register to reply

Related Discussions
Kolmogorov-Smirnov and P-value Set Theory, Logic, Probability, Statistics 11
math to compute width of gaussian PDF(x) given sigma & N values of random variable x Set Theory, Logic, Probability, Statistics 10
Question about the Nagata-Smirnov metrization theorem Differential Geometry 11
Broken ties General Discussion 14
Kolmogorv-Smirnov goodness of fit test Calculus & Beyond Homework 2