10 Golden Minutes

1
10 Golden Minutes

• Study power
• Joe West

2
Research question

• Will the audience fall asleep more during this
  talk than during the data management talk that
  Kate is going to give tomorrow?
• Proxy measure for boredom
• Unexposed group my talk
• Exposed group (or intervention) Kates talk

3
Outcome

• Falling asleep
• Proportion that meet the outcome in the unexposed
  group expected to be 80 (anecdotal experience)

4
Hypothesis

• Null hypothesis is that the proportion who will
  fall asleep in this talk is the same as in Kates
  talk (i.e. 80)
• Alternative hypothesis is that 80 will fall
  asleep during this talk while only 70 will fall
  asleep during Kates talk

5

• A difference, to be a difference, must make a
  difference
• Gertrude Stein (1874-1946) Writer , eccentric and
  self-styled genius

6
Significance

• Important clinical difference is 10
• I want to sure that only 5 of the time will I
  mistakenly reject the null hypothesis
• i.e. actually both talks are boring, neither is
  more boring than the other, but we find a
  difference
• Alpha0.05 this will occur 1 in 20 times

7
Power

• What is power?
• If there is actually a difference, with what
  probability would you like to detect this
  difference?
• I want to be 90 sure that I have not mistakenly
  accepted the null hypothesis
• i.e. actually finding the important news that I
  am more boring than Kate

8
Calculation

• P1 - asleep in my talk 80
• P2 - asleep in debate 70
• Difference 10
• Alpha 0.05
• Beta 0.1 i.e. 90 power
• N 412 per group

9

• http//www.stat.uiowa.edu/rlenth/Power/

Slide 3