当前位置:网站首页>Graphical FDR

Graphical FDR

2021-06-24 00:55:22 Zhang Chuncheng

The illustration FDR

Since writing is not seen , So I decided to change it to a picture . This series strives for a picture to tell a story . This picture is right FDR The calibration process and principle are illustrated .

The following brief explanation , You can choose to see it or not , Because it's to make up for 300 The minimum number of words required , No effect at all .


FDR Principle diagram

FDR.png
FDR.png

This picture attempts to illustrate a sentence :

according to FDR Methods after statistical test and correction , The expected upper limit of the first type error ratio is not greater than the given value .

This picture is from the classical literature Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing[1].

A different approach to problems of multiple significance testing is presented. It calls for controlling the expected proportion of falsely rejected hypotheses - the false discovery rate.

Brief description

So-called False Discovery Rate, It is to limit the proportion of false rejections of null hypothesis . In the picture , Corresponding to random variables .

Its practical significance lies in avoiding multiple comparison risks

  • When not checked , Because through repeated comparisons , It is bound to cause a certain degree of “ False alarm ”;
  • After verification , You can reduce that risk ;
  • In numerical terms , Samples that pass the verification , Its “ False alarm ” There will be a strict upper limit on the proportion , So as to achieve risk control .

meanwhile , You can see , Compared to the extremely strict Verification method ,FDR Verification is a very loose way of verification .

Reference material

[1]

Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing: http://engr.case.edu/ray_soumya/mlrg/controlling_fdr_benjamini95.pdf

版权声明
本文为[Zhang Chuncheng]所创,转载请带上原文链接,感谢
https://chowdera.com/2021/06/20210624005412423k.html

随机推荐