L22-23: Chance

Question 1

What is chance?

Answer 1

Sampling error - form of random error

Question 2

How do we reduce chance/random error from occurring?

Answer 2

By increasing sample size

Question 3

How does increasing the sample size reduce effects of chance?

Answer 3

• increases likelihood of getting a more representative sample
• improves precision of the parameter

Question 4

How is chance measured? What is the basis of clinical/statistical significance in each of these measures?

Answer 4

Confidence intervals: can measure clinical significance.
- if CI contains null value, finding is not clinically significant (chance likely an explanation)
-if CI does not contain null value, finding is clinically significant (chance is not a likely explanation)

p-values: can measure statistical significance (basis is 0.05) => measures probability of getting a study estimate
- if p-value < 0.05, finding is statistically significant
- if p-value > 0.05, finding is not statistically significant

Question 5

If the parameter of the p-value contains the null value, what does this mean? and vice versa?

Answer 5

If it contains the null value, then there is no association of exposure and risk of disease in the population.

If it doesn’t contain the null value, then there is an association present.

Question 6

What are the null values for Relative Risk and Risk Difference?

Answer 6

RR = 1, RD = 0

Question 7

What are the 2 errors that can be produced with p-values?

Answer 7

Because p-values can determine association, this may sometimes result in misclassification.

Type 1: when there is an association, but shouldn’t be
Type 2: when there isn’t an association, but should be

Question 8

Type 2 errors are commonly due to what?

Answer 8

Small study sample

Small study sample => larger p-value

Question 9

What are the limitations of p-values?

Answer 9

• don’t indicate clinical significance
• they are arbitrary thresholds meaning they’re usually based on random choice
• only about null value