发帖
每发1个新帖
可以获得0.5个丁当奖励
回帖

【原创】Power/sample size analysis: step-by-step [精华]

楼主

丁香园中级站友

+1 积分
1楼

这个帖子发布于9年零193天前，其中的信息可能已发生改变或有所发展。

I saw a lot of guys here raised questions on power/sample size analysis. Power/sample size is the priority issue of genetic association study. It's quite important but it's no so complicated as thought. Many software can handle this. The best one is nQuery Advisor. It's quite professional and powerful. But it's too expensive. So free small ones should be our favorite.

Too many free software can do power analysis. They all use the same or similar statistical models so their outcomes are also similar. Here I will introduce a software "PS: Power and Sample Size Calculation" as a example to briefly show how to calculate statistical power or sample size for association study. This software has several pros, such as small size, windows-based interface (most operations are based on mouse clicks), easily accessed help etc. Our lab has always been using this software for more than 5 years and published more than 10 association study publications. The reviewer don't care what software you use, they only care what the results stand for. I seldom use other software, but I think most of them are similar. The difference may only be the data input method and interface.

Home page: http://biostat.mc.vanderbilt.edu/wiki/Main/PowerSampleSize
Remember, if you use this software in your manuscript, you'd better cite the authors' publication(s). It's respect to the authors for their generousness and free distribution.

Let's start:
1) download and install the software. After the installation, program shortcut will appear in the Program menu.

2) Click the shortcut, you can see the software main interface.

3) Click the tag lag "Dichotomous", here is what we need for chi-square test. If you are running other test (e.g. T-test, regression etc), you can click other tags interested.

4) Choose what test you want, sample size or power

I use power as an example.

5) Choose whether your design is matched or independent

Then choose whether it's a case-control or prospective study

If you choose case-control, there will be two additional menus
One is to choose how your hypothesis express, odds-ratio or two proportions (MAF in case and control group)

Another is to choose whether your study is corrected, usually we select uncorrected chi-square because each cell is definitely larger than 5.

6) Then the population parameters input boxes show up. You can click each box's name (blue underlined character), the help document will appear to interpret what the parameters stand for.