Bootstrapping and p-values (t-test when using bootci)

Question

Kilin 2016-2-4

2
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/266682-bootstrapping-and-p-values-t-test-when-using-bootci

回答： John Williams 2018-10-15

Hello,

I have to report some results from random experiments, and I need to report confidence intervals and p-values as well. I am not very familiar with statistics, so I hope you can help me with this.

Given that I can't assume that my distribution is normally distributed, and that I don't have many samples, I've been suggested to adopt a bootstrapping method.

     RESAMPLING_SIZE = 10000;
     CI = bootci(RESAMPLING_SIZE, {@mean, myData});

I then plot mean(myData) along with CI.

First question: Is it correct to plot mean(myData), or I should plot an overal mean as estimated by the bootstrapping method?

Second question: say I want to compare results from two different experiments, and I have already applied bootstrapping on the two sets of data. How to compute p-values? Can I use ttest or ttest2 in combination with bootci?

Thank you very much

Best regards

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

John Williams 2018-10-15

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/266682-bootstrapping-and-p-values-t-test-when-using-bootci#answer_341538

While this post is 2 years old, just in case it helps someone - here is some code that can compute bootstrap CIs and P values based on some precomputed resample and leave one out statistics.

https://www.mathworks.com/matlabcentral/fileexchange/69119-bca_bootstrap

The methods are described in the eAppendix of this paper:

https://www.ncbi.nlm.nih.gov/pubmed/19581566