Choosing the right statistical test depends on three things:
- Data type (parametric vs non-parametric)
- Number of groups
- Paired vs unpaired
Parametric vs Non-Parametric
Parametric tests = interval or ratio data (normally distributed)
Non-parametric tests = ordinal or nominal data (or not normally distributed)
Two Groups – What Test Do You Use?
Paired data (same group measured twice):
- Parametric: Paired t-test
- Non-parametric: Wilcoxon signed-rank test
Unpaired data (independent groups):
- Parametric: Unpaired t-test
- Non-parametric: Mann–Whitney U test
Three or More Groups – What Test Do You Use?
Paired data (e.g. same subjects under three conditions):
- Parametric: Repeated Measures ANOVA
- Non-parametric: Friedman test
Unpaired data (e.g. 3+ different groups):
- Parametric: One-way ANOVA
- Non-parametric: Kruskal–Wallis test
📉 Comparing Proportions or Percentages
Paired data (2×2 table):
- Small or large sample: McNemar’s test
Unpaired data:
- Small sample: Fisher’s Exact test
- Large sample: Chi-squared test
The Null Hypothesis
Says there’s no true difference between groups
Your goal in most studies is to reject it
Understanding the p-value
This is the chance of finding a result as extreme as yours, assuming the null hypothesis is true
p < 0.05 = statistically significant
Depends on sample size and effect size
❌ Type I vs Type II Errors
Type I error (α): False positive:
- You reject the null hypothesis, but it was actually true
- Controlled by the p-value
Type II error (β): False negative:
- You fail to reject the null, but it was actually false
- More likely if your sample size is too small
Power of a Study
Power = 1 – β
It’s the probability of correctly rejecting a false null hypothesis
High power = more likely to detect a real difference if one exists
Z-Scores – How Far From the Mean?
Z = number of standard deviations from the mean
Z = 0 → exactly average
Z > 2.2 → statistically significant (reject the null)
Key Exam Tips
Always check paired vs unpaired, parametric vs non-parametric
Don’t confuse p-values with clinical significance
Know your error types: Type I = false alarm, Type II = missed signal
Power matters – especially in negative studies
Z-scores test how extreme your result is compared to the population
Thank you to the joint editorial team of MRCEM Exam Prep for this ‘Exam Tips’ blog post.
