huangyhg

kolmogorov - simrnov test
hi all,
i have done some measurments and want to fit these meatsurments to certian distributions , read about ks test but cant understand how to chose the constant value that if the maximum differnece d between the know distribtuin and the function under test is less than or equal to this constant then the hypothesis is right , other wise it is not how can i know what constant to chose and how can i know according to the constant what value of confidence level is assumed ,,,
thnaks for any help in advance
no answer plz answer me
there's no such test. accuracy is important.
hi greg , i read this information i told you in the book , degroot " probability and statitics " , and it is like this how it works in this book , i am still confused
you find the constant, the critical value, from tables if your program does not generate it. you have to chose the kind of your test: single sided or double sided, you chose the significance level and you have the number of points from your measurements.look at www wikipedia.
m777182
saw the website and it is not related , can you give it to me again
wadah
perhaps your book is in error or you are misinterpreting the information; since there are other sites that reflect the same material that greg's link presents:
1. first you have to sort your data in ascending order
2. set up table:
i y(i) fn(yi) f(yi) dn+ dn-
where i is an index 1,2,3...n
where y(i) is your sample data
fn(yi) = i/n
f(yi) is your function for y(i) for the test distribution
dn+ is absolute value of [fn(yi) - f(yi)]
dn- is absoulte value of [f(yi) - (i-1)/n]
3. your test statistic dn = max{dn+,dn-}
compare to a table value and reject ho if dn > d
thnaks aalot penn state ie

__________________
借用达朗贝尔的名言：前进吧，你会得到信心!
[url="http://www.dimcax.com"]几何尺寸与公差标准[/url]