Biostats lecture 4

5 questions to ask for determining type of statistical test

1. what is the scale of measurement?

2. how many samples

3. independent or dependent samples?

4. (only for continuous variables) parametric vs nonparametric

5. (only for nominal vairables, independent samples) chi square vs fishers

what is the scale of measurement

• nominal (categorical)

• ordinal (categorical)

• interval/ration (continuous)

how many samples?

• two vs more than 2

group

• treatment groups, unique individuals receiving the same treatment

sample

• subset from a population, could be measures or persons, may be measures of the same group of individuals

independent/unrelated samples

• randomized to groups

dependent/related samples

• repeared measures: same group observed multiple times

• matched groups: patient pairs matched on some characteristics (gender, race/ethinicity)

repeated measures

• type of dependent sample

• same group observed multiple times

• ex: before treatment, 1,3,6,12 months after treatment)

parametric vs nonparametric

• can i assume normal distribution

• if the distribution is unknown, is the sample size large?

can i assume normal distribution

• is the distribution approximately normal or skewed

• if skewed distribution, nonparametric

• otherwise parametric

nonparametric

• use if distribution is skewed

if the distribution is unkown, is the sample size large?

• small sample size (<30) AND unknown distribtion→ nonparametric

• otherwise parametric

chi square vs fishers

• only for nominal variable, independent samples

• small study → total number < 30

• OR

• any category < 10 (conservative rule of thumb)

• if yes to one or both → fishers

• otherwise chi square

fisher’s

• nominal variable, independent samples

• small study total number < 30

• or

• any category < 10