Hypothesis testing

Question 1

What is a ‘hypothesis’?

Answer 1

A hypothesis is a testable statement which entails our beliefs about our observations

Question 2

What is a null hypothesis?

Answer 2

It is called ‘null’ because essentially, it states that something ‘equals zero’, or in other words there is no finding

Question 3

What is an alternative hypothesis?

Answer 3

It is called ‘alternative’ because it is complementary to the ‘null’. It is essentially saying there will be a difference

Question 4

What is a one-sided hypothesis and a two-sided hypothesis?

Answer 4

One-sided is a prediction it will go one way e.g. up

Two-sided is a prediction it could change in either direction

Question 5

When writing a hypothesis what are the 2 questions we should ask ourselves?

Answer 5

1. How will we be measuring it? e.g. cm, lbs

2. Which parameter will we be measuring? e.g. mean, medium

Question 6

If we do not find a difference, what do we report?

Answer 6

We never accept the null hypothesis, we say we have not found enough evidence to reject it or we have failed to reject the null hypothesis

Question 7

What is a type 1 error? What is this conventionally set too?

Answer 7

To think you have found a difference, but you haven’t, the data was not true or fake.
By convention α is set to 0.05, that is: if I say that there is a difference in the population I expect to be at least 95% confident about my result.

Question 8

What is a type 2 error? What is this conventionally set too?

Answer 8

To think that there is no difference, but in fact there was one.
By convention set to 0.80, that is: if there was a difference in the population I expect to be at least 80% confident that I would be able to reveal it

Question 9

What symbol would you use to symbolise a type 1 error?

Answer 9

Alpha symbol (significance level alpha)

Question 10

What symbol would you use to symbolise a type 2 error?

Answer 10

Beta symbol (Power B-1)

Question 11

When rejecting the null hypothesis, we have ….. chance of finding a difference is the sample size is large.
We have ……. chance to reject the null hypothesis if the difference is larger

Answer 11

Better
Better

If the null hypothesis is false (there is a treasure), my power to correctly reject it (1-β) is larger if the treasure is big or if I sample more areas.

Question 12

What are the 4 parts of the objective power analysis?

Answer 12

Power 0.80
Sample Size
Error allowed a=0.05
Effect size

Question 13

What is the objective power analysis?

Answer 13

It is the process of working out 1) if your sample is big enough to find a difference and 2) how much power my experiment has to reveal a difference