Topic 11: statistical hypothesis testing

Question 1

Give the 2 possibilities of association between 2 variables

Answer 1

1) Association doesn’t exist in population = variables not linked
2) Association exists in population = variables linked

Question 2

Define null hypothesis

Answer 2

• No association between variables in population
• Use statistical analysis to decide if there is evidence to reject null

Question 3

Define alternative hypothesis

Answer 3

• There is assciation between 2 variables in population

Question 4

Explain the process of hypothesis testing

Answer 4

• Define null/alternative hypotheses
• Assume no association = initially null
• Define what is sufficient evidence against null = significance
• Collect data = see if provides sufficient evidence against null
• Calculate value of test statistic = derive probability against null = p-value
• Interpret p-value as significance level

Question 5

Describe p-value

Answer 5

• What is probability of getting an association as strong as observed in sample if there is no association in source population = see association by chance even with no association
• Low p-value = unlikely association in sample by chance = if no true association existed
• Lower = reject null = favor alternative = association exists

Question 6

Define significance value

Answer 6

• Significance level = 5%
• P-value = <0.05 = value statistically significant
• P-value = ≥0.05 = not SS = cannot reject null

Question 7

Give assumptions in presence of association

Answer 7

1) Large sample size = smaller p-value
2) Large magnitude = smaller p-value

Question 8

Describe 95% CI rejecting null for mean difference

Answer 8

• If 95% Cl = 0 = null cannot be rejected
• 0 = no difference between 2 means
• 0 = likely value in source population

Question 9

Describe 95% CI rejecting null for regression/correlation coefficient

Answer 9

• If 95% Cls = 0 = nulll cannot be rejected
• 0 = no correlation between 2 variables + slope of 0
• 0 = likely value in source population

Question 10

Describe 95% CI rejecting null for odds/risk/rate ratio

Answer 10

• If 95% Cl = 1 = null cannot be rejected
• 1 = equal risks/rates/odds between both groups
• 1 = likely value in source population