1
$\begingroup$

I need to calculate the coefficient of the Kolmogorov-Smirnov test for any given null hypothesis rejection level.

For example, have a look at the table in this wikipedia entry. It only gives $c(\alpha)$ for $\alpha=0.10, 0.05, 0.025, 0.01, 0.005, 0.001$. How can I calculate $c(\alpha)$ for a wider range of $\alpha$s?

If it is not trivial, can any one guide me on a place I can find a table for $c(\alpha)$ with larger $\alpha$ values?

Thanks in advance.

$\endgroup$
3
  • $\begingroup$ Normally people use programs which can supply p-values. Note also that for small samples, those values in that table are only approximate. You may find this answer of some value: How do I find values not given in (/interpolate in) statistical tables?. Alternatively, this might be of some value. $\endgroup$
    – Glen_b
    Commented Nov 5, 2013 at 10:55
  • $\begingroup$ Thank you very much for the references, I will definitely take a good look through them to learn more. I think I found my answer in the answer to this question: stats.stackexchange.com/questions/13709/… (Marsaglia, et al. (2003)). I read it and found a nice C code (that is the language I am working in) to give me the probability. It works fast and nice. $\endgroup$
    – astroboy
    Commented Nov 5, 2013 at 11:13
  • $\begingroup$ An alternative (and more easily accessible) link to the original paper should be here. Note that the tables given there are quite extensive. The $L(z)$ values correspond to $(1-\alpha)$ and the $z$ to $c(\alpha)$. $\endgroup$
    – Glen_b
    Commented Nov 5, 2013 at 11:17

1 Answer 1

2
$\begingroup$

I found this paper (Marsaglia, et al. (2003)), that has a method to approximate the probability of obtaining a certain $D_n$ for a certain number of data. It Also has a C program that I used in my program and worked excellently.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.