parametric bootstrap for kolmogorov smirnov test

Hi,

I want to perform KS test for my sample data and test it using parametric bootstrap of KS test and to get a p-value. In my case the empirical distribution is using Weibull's plotting position formula fitted to the data, which i have already written using sorted data points. But in MATLAB it uses its own empirical cdf function. and I want to write my own function, kindly let me know how i can proceed.

Answer 1

[the cyclist]

It sounds to me like you can use the syntax

 >> h = kstest(x,cdf)

where "cdf" is a two-column matrix. The first column of cdf is a list of possible x values, and the second column is the value of your theoretical cdf at those values of x.

Read more at

 >> doc kstest

Does that help?

[Pg]

But as far I know it only test data with respect to normal distribution, I h've to check for various other distribution. More over I'm using Weibull's plotting position formula for to compare empirical CDF. It uses function ecdf to test with theoretical distribution. So I want to write my own function.

[the cyclist]

When kstest() is called with only one input argument, it will assume that you want to test against a normal distribution. But with the syntax I suggest, you will enter your own array that defines the CDF to be whatever you want. You might want to make a more careful read of the documentation: http://www.mathworks.com/help/toolbox/stats/kstest.html.

[Pg]

Ok That I was unaware, thanks for that

Answer 2

[Francesco Falcone]

Do you know if the kstest automatically performs a parameteric bootstrap? I thought it needed to be expanded in that case. Having the same problem so that's why I'm asking.

[the cyclist]

I don't fully understand this question. I suggest you submit a brand-new question (with a bit more description), rather than appending it as an "answer" to a 5-year-old question. (It was random chance that I ran across this.)