How Can I draw confidence intervals on probability plot?

13 次查看(过去 30 天)
Hello I would like to draw probability plot using the probplot function for different distributions (e.g. lognormal and weibull) and i would like to know how i can draw the confidence intervals so that i can check if my data points fit into the confidence bounds. Please note that I have censored and uncensored data so I am forced to use probplot in order to take the censored data into account. I also tried to used the "dfittool" in the Statistics toolbox but even there I could not find the way to draw confidence bounds.
Another question is how to determine the p-value of the probability plot so that I can see how good my data fit the chosen distribution.
Thanks
Paolo

回答(4 个)

Star Strider
Star Strider 2012-10-19
You did not mention if you had used edcf in your analysis.
Would it do what you want?
  1 个评论
Star Strider
Star Strider 2012-10-19
I'm still not certain what you want, but the function that may be closest to what I believe you're looking for is probably the Kolmogorov-Smirnov test. I'm not sure you can plot its confidence limits, but it may help you decide the distribution that most closely fits your data.
The only other possibility I can think of is for you to use nlinfit and its related functions to fit your data (using an anonymous function to code for the relevant distributions), and then use nlpredci to plot the errors of the fit. If you have the errors and you simply want to plot them, consider errorbar.

请先登录,再进行评论。


Paolo  Bortot
Paolo Bortot 2012-10-19
Thanks for your prompt reply. Actually this was very helpful but it's not what I meant.
I am trying to understand which is the distribution (either weibull or lognormal) that best fits my data and I wanted to draw the condidence bounds on the same plot with my probability plot. How do I do that?
Thanks
Paolo

Paolo  Bortot
Paolo Bortot 2012-10-22
Thanks for the answer.
I relaized the question was a bit unclear, I try to make myself clearer.
This the problem:
I have a set of experimental data that I presume, most likely, will be distributed according to either a Weibull or a lognormal distribution. I use probplot to look at the two probability plots. Now the questions are: 1) Can I plot on the same graph the confidence lines and not just the line representing the probability plot? 2) What is the best way to see how good my data fit to either weibull or lognormal? In other words, how do I do a p-value test for the two distributions?

Sachin Ganjare
Sachin Ganjare 2012-10-22

标签

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by