Modifying the Parameter Estimation Method of wblfit Function

Question

Ali 2015-3-27

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/194711-modifying-the-parameter-estimation-method-of-wblfit-function

回答： Dutchie 2016-11-1

By default, the wblfit function returns the 'maximum likelihood estimate' as describied below:

parmhat = wblfit(data) returns the maximum likelihood estimates , parmhat, of the parameters of the Weibull distribution given the values in the vector data, which must be positive. parmhat is a two-element row vector: parmhat(1) estimates the Weibull parameter a, and parmhat(2) estimates the Weibull parameter b, in the pdf.

How can I change the parameter estimation method from maximum likelihood estimate to rank regression method of the least square curve fit? I've spent 2 days trying to figure this out and have had no luck. The 'rank regression method' is very common when it comes to weibull analysis on small sample sizes. Any help would be greatly appreciated.

<http://www.mathworks.com/help/stats/wblfit.html>

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Tom Lane 2015-4-4

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/194711-modifying-the-parameter-estimation-method-of-wblfit-function#answer_173934

在 MATLAB Online 中打开

I think you should not modify wblfit, but just do it directly or write a separate function. Here's an example of how to do what I think you want:

% Random sample with paramaters 1000 and 1.3
n = 1000;
x = wblrnd(1000,1.3,n,1);
% Fit by maximum likelihood
p = wblfit(x)
% Create probability plot on log scale
z = wblinv(((1:n)-.5)/n,1,1)'; % idealized sample from standard Weibull
plot(log(z),sort(log(x)),'bx')
% Fit points on plot by least squares
b = polyfit(log(z),sort(log(x)),1)
line(log(z),b(1)*log(z)+b(2),'LineStyle',':')
% Transform to Weibull parameter scale
p1 = 1/b(1)
p2 = exp(b(2))

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

Ali 2015-4-5

Tom,

Thank you for your feedback! The result was close but not exactly the right answer.

For a sample of failure times of [30 49 82 90 96]. The Eta would be 79.8 and the Beta would be 2.165 using the rank regression model. Using your algorithm, I arrive at an Eta 78.723 and a Beta of 2.465, which is very close.

I am trying to go through and determine if I can make the necessary adjustments to arrive at the proper solution.

请先登录，再进行评论。