h***s
发帖数: 35 | 1 Hello,
Very appreciate in advance.
Andseron-darling test defintion: integral(Fn-F(x))/(F(x)*(1-F(x))dF, Fn is
empirical data, F(x) is CDF of thoery distribution.
How to get its practical test formula: -n -sum((2k-1)/n*(log(Yk)+log(Y(n-k+1
))?
I read his orginal paper "A Test of Goodness of Fit", it only mentions to
divide into multiple integral part: (-infinity, x1), (x1, x2)... (xn, +
infinity), for each part do integration. But I cannot get -n -sum((2k-1)/n*(
log(Yk)+log(Y(n-k+1)).
Does anyone has idea? | l******n
发帖数: 9344 | 2 直接分段积分，大一习题呀
+1
*(
【在 h***s 的大作中提到】
: Hello,
: Very appreciate in advance.
: Andseron-darling test defintion: integral(Fn-F(x))/(F(x)*(1-F(x))dF, Fn is
: empirical data, F(x) is CDF of thoery distribution.
: How to get its practical test formula: -n -sum((2k-1)/n*(log(Yk)+log(Y(n-k+1
: ))?
: I read his orginal paper "A Test of Goodness of Fit", it only mentions to
: divide into multiple integral part: (-infinity, x1), (x1, x2)... (xn, +
: infinity), for each part do integration. But I cannot get -n -sum((2k-1)/n*(
: log(Yk)+log(Y(n-k+1)).
| h***s
发帖数: 35 | 3 Yes, I made a mistake. in [x1, x2), the empirical propability should be same
1/n for x1 and x2, insread of 1/n for x1 and 2/n for x2. Thank you very
much! |
|
