3

I am working on a hypothesis test and need to separate a group of people into a control and a test group. What is the best way to calculate that, if I have N people?

mdewey
  • 16,541
  • 22
  • 30
  • 57
jeangelj
  • 393
  • 2
  • 5
  • 14
  • 1
    You're concerned with a concept called "statistical power". Google it. Then have a look at the answers to this question: http://stats.stackexchange.com/questions/54693/power-for-experimental-design/54717#54717 – generic_user Mar 03 '17 at 15:49

1 Answers1

6

You'll get the most power with a 50:50 split --- equal numbers of people getting the treatment and control.

The only reason not to use an equal split would be if the treatment is expensive, rare, time-consuming to prepare, risky, etc... In those cases, you might prefer to give the treatment to fewer people than the control.

Harvey Motulsky
  • 14,903
  • 11
  • 51
  • 98
  • Thank you - how do I calculate the power? – jeangelj Mar 03 '17 at 15:36
  • 1
    @jeangelj . This site can't serve as a statistics textbook. There are lots of books, and web pages, that explain how to compute sample size. Read these, then write back to this site if you are confused about something. – Harvey Motulsky Mar 03 '17 at 17:30