Lecture 3: Hypothesis Testing

Hypothesis (H1)

A testable statement about the characteristics of the population

Null Hypothesis (H0)

No association or difference

Study designs for comparing two groups

Paired-design or two independent sample design

Paired-design study

Participants are grouped into pairs based on shared characteristics relevant to the study.

What type of data does paired-design study?

Before-and-after and matched-case-control

Matched-case-control

Cases are matched with controls based on shared characteristics to determine the association between exposure and outcome

Confidence Interval Approach

Draw a conclusion about a population parameter

E(X)

Expected value of X can sometimes be used to denote the mean

µ

expected population mean change

How is the confidence interval approach used?

Creates a range of possible values for μ using data from a single paired sample

Sampling distribution

theoretical distribution of all possible values of the mean from samples in the same size

P-value (0-1)

The probability of achieving the test statistic as/more extreme than observed if null hypothesis is true

Small p-value (<0.05)

Incompatible with null hypothesis → reject

Confidence interval does not include 0

More likely to be statistically significant

Type I error

Alpha level (reject null, null is true)

Type II error

beta (accept null, alternative is true)