Week 9

Question 1

What is type 1 error? (3)

Answer 1

Rejecting the null hypothesis when it is correct

‘False positive’ leading to too many positive results

Question 2

What is type 2 error? (3)

Answer 2

Accepting the null hypothesis when it is incorrect

‘False negative’ leading to too many negative results

Question 3

How is type 1 error denoted and how can it be reduced?

Answer 3

Denoted as α, usually set to 0.05
It is ok to have 5% probability of incorrectly rejecting true null hypothesis

Reduced by:
Lower α value

Question 4

How is type 2 error denoted and how can it be reduced?

Answer 4

Denoted as β
Power of a test is 1- β

Reduced by:
Large sample size
Larger effect size

Question 5

How are type 1 and type 2 errors related to one another?

Answer 5

Errors are inversely proportional

Relaxing the alpha level (eg: p= 0.1) will increase type 1 error, and give more false positives but less type 2 errors (false negatives)

Tightening the alpha level (eg: p= 0.01) will decrease type 1 false positive errors but give more type 2 false negative errors

Question 6

What is effect size (2)

Answer 6

The difference in value of the primary endpoint between the test group and control group
The larger the expected effect size, the smaller the size of the sample required to detect it

Question 7

What factors should be looked into when deciding sample size?

Answer 7

Big enough for:
Minimising type 2 errors
Represent population so result is generalisable

Small enough to:
Be feasible and achievable
Minimise risk to participants, cost and waste

Question 8

What factors are required for calculation of sample size? (4)

Answer 8

• Analysis plan
• To set α and 1-β (usually 0.05 and 0.8)
• Effect size
• Population variance