In: Statistics and Probability
A pharmaceutical company develops drugs to increase apoptosis in cancer cells. Each candidate drug is tested in mice, using an experiment design that has 90% power at a = 1%. 1) For every 1000 drugs that have no real effect, about how many will the company mistakenly conclude work? 2) For every 1000 drugs that actually work, about how many will the company correctly conclude work?
Null Hypothesis H0 is that the drugs have no real effect.
Alternative Hypothesis H1 is that the drugs actually work to increase apoptosis in cancer cells.
Probability that when the drugs that have no real effect, the company will mistakenly conclude that the drugs work
= Probability that H1 is accepted to be true when in reality H0 is true
= Probability of Type I error = a = 1% = 0.01
Thus for every 1000 drugs that have no real effect, the company will mistakenly conclude that 0.011000 = 10 drugs work.
Probability that when the drugs actually work, the company will correctly conclude that the drugs work
= Probability that H1 is accepted to be true when in reality H1 is true
= 1 - Probability that H0 is accepted to be true when in reality H1 is true
= 1 - Probability of Type II error = Power = 90% = 0.9
Thus for every 1000 drugs that actually work, the company will correctly conclude that 0.91000 = 900 drugs work.