Chapter 6

Question 1

Name 3 ways that null hypothesis testing relies on size of sample

Answer 1

• Non-significance can be due to small sample size
• Small sample size leads to smaller power in tests (bigger chances of making a type 2 error)
• Practically irrelevant tiny effects can be statistically significant in large samples

Question 2

Important to keep in mind about statistical significance

Answer 2

It is not a measure of strength or importance, it ONLY means that H0 must be rejected

Question 3

When should we use effect size?

Answer 3

To say something about the magnitude of an effect in the population

Question 4

What does it mean when we have low test power?

Answer 4

-There can be substantive differences between true and hypothesized population values even if the test is statistically significant
- With low power, difference has to be quite large to obtain statistically significant results but
we are more likely to have an effect that is practically relevant

Question 5

What is knocking down straw man?

Answer 5

Straw man is testing with the same data over and over again. Knocking down straw man is the correction.

Question 6

Alternatives to null hypothesis testing

Answer 6

Estimation instead of hypothesis testing: reporting & interpreting confidence intervals rather than relying solely on H0 testing

Question 7

What is a meta-analysis?

Answer 7

collecting previous studies on same topic & combining results of the studies making one large sample out of small samples

Question 8

Bayesian inference

Answer 8

Previous knowledge is starting point
- Sample that we draw is a means to update the knowledge we already have
- Frequentist inference = no true population value, population value is a random value (has a
probability)
- Credible interval = parameter has a probability, thus, it can lie within the interval with 95%
probability