IB Biology IA Statistical Tests: t-test vs ANOVA vs Chi-squared (2025 Rubric)

Your statistical test is worth 1–2 marks on the Data Analysis criterion (6 marks total). Pick the wrong test and your Conclusion will wobble too, because examiners read both criteria together.

This guide walks you through the decision tree from data type → test choice → assumption check, and shows you exactly what the 2025 IB Biology IA rubric rewards.

The 5-Second Decision Tree

If you can identify your row in this table, you are already at a 5/6 Data Analysis floor.

When to Use Each Test — With IB Biology Examples

t-test (two means)

Use when you are comparing the means of exactly two groups of continuous data.

IB Biology example: Effect of light vs dark on Elodea photosynthesis rate. You have 10 replicates in light, 10 replicates in dark, and you want to know whether the mean rates differ significantly.

• Independent t-test: different subjects in each group.
• Paired t-test: same subject measured twice (before/after).

Assumptions to check:

1. Data is approximately normal (Shapiro-Wilk test or Q-Q plot).
2. Equal variances (Levene's test) — if unequal, use Welch's t-test.

2025 rubric signal: Examiners reward stating the p-value, the 95% confidence interval for the mean difference, and Cohen's d (effect size). A p-value alone caps you at 4/6.

One-way ANOVA (3+ means)

Use when comparing three or more group means. The single biggest error on 2025 IAs is using a t-test on 4 groups — this is mathematically invalid because it inflates your Type I error rate.

IB Biology example: Effect of 5 different enzyme pH levels (3, 5, 7, 9, 11) on amylase activity. You have 5 replicates per pH, 25 data points total.

Assumptions:

1. Normality within each group.
2. Homogeneity of variance across groups (Levene's).
3. Independence of observations.

2025 rubric signal: If you write "ANOVA showed significant differences (F = 12.3, p < 0.001), and Tukey post-hoc identified pH 7 as significantly higher than all others" — that is 5/6 territory. Adding effect size (η² or ω²) is the 6/6 move.

Chi-squared (categorical data)

Use for counts, not means. Two common flavours:

• Goodness of fit: Compare observed frequencies to expected. Classic use: testing Mendelian 3:1 or 9:3:3:1 ratios in a PTC tasting or Drosophila IA.
• Independence: Test whether two categorical variables are associated. E.g. smoker vs non-smoker × lung infection rate.

IB Biology example: You observe 78 tasters and 22 non-tasters for PTC in 100 volunteers. Hardy-Weinberg expected under p² + 2pq + q² = 1 predicts different frequencies. Chi-squared goodness of fit tests whether the deviation is significant.

Assumption: Expected count ≥ 5 in each category. If any cell has expected < 5, use Fisher's exact test instead.

2025 rubric signal: Include the chi-squared statistic, degrees of freedom, and p-value. State the null hypothesis explicitly.

Pearson (linear correlation)

Use for two continuous variables where you expect a linear relationship.

IB Biology example: Temperature (continuous) vs enzyme initial reaction rate (continuous). You have one data point per temperature but 5 temperatures.

Assumption: Linear relationship. If you see a curve (e.g. bell-shaped enzyme temperature response), regression with a non-linear model is better.

2025 rubric signal: Report r, r², and the p-value. Explain what r² tells you about the proportion of variance explained.

Spearman (rank correlation)

Use when one or both variables are ordinal, or when the relationship is monotonic but not linear.

IB Biology example: Ranked species abundance across 5 habitats. Use Spearman.

Common Mistakes That Drop 2 Marks

1. Using a t-test on 4 groups

This is the single most common statistical error in IB Biology IAs. It inflates your false positive rate. The fix: one-way ANOVA with Tukey HSD post-hoc.

2. Not checking normality

If you run a t-test on a heavily skewed dataset, the p-value is unreliable. Always check (Shapiro-Wilk on each group, or a Q-Q plot). If non-normal, use the non-parametric equivalent (Mann-Whitney U).

3. Reporting only the p-value

p < 0.05 is not enough. Examiners expect effect size and confidence interval. For ANOVA, report η² (proportion of variance explained). For t-tests, report Cohen's d. For correlations, r² does the same job.

4. Using chi-squared when you should use Fisher's exact

If any cell has an expected count below 5, chi-squared becomes unreliable. Example: you have 100 volunteers split across 8 blood-group categories — several cells will have expected counts of 1-2. Use Fisher's exact test instead.

5. Picking the test after seeing the data

Examiners flag this. Your test should be justified in Research Design based on the planned experiment, not chosen post-hoc to make your data look significant.

What Examiners Reward Under Each Criterion

Worked Example — A Grade-7 IA

Research Question: How does temperature (10, 20, 30, 40, 50 °C) affect the initial rate of catalase-mediated O₂ evolution from potato extract?

Data: 5 temperatures × 6 replicates = 30 data points. Initial rates in cm³ s⁻¹.

Test selection rationale (Research Design):

"Because I have five independent groups (temperatures) of continuous data (initial rate), the most appropriate test is one-way ANOVA with Tukey HSD post-hoc to identify which temperature pairs differ. I will verify normality (Shapiro-Wilk at α = 0.05 per group) and homogeneity of variance (Levene's test) before applying the parametric test. If either assumption is violated, I will use Kruskal-Wallis with Dunn's post-hoc instead."

Result reporting (Data Analysis):

"ANOVA showed a significant main effect of temperature on initial rate, F(4, 25) = 42.3, p < 0.001, η² = 0.87 (large effect size). Tukey HSD identified 40 °C as significantly faster than 10, 20, and 50 °C (all p < 0.01). Rate at 40 °C: mean = 2.4 ± 0.18 cm³ s⁻¹ (SE)."

That single paragraph is a 6/6 Data Analysis mark.

Next Steps

• Already chose an IA topic? Start with the IA Guide (2025) for the full 4-criterion breakdown.
• Not sure which topic? Browse 50+ IA topic ideas by theme.
• IA broken and deadline close? The IA troubleshooting guide covers the 12 most common failure modes.

Not sure which test fits your data?

If you have your IA data collected but are still deciding between a t-test, ANOVA, or Chi-squared, message us on WhatsApp at +91 88264 44334 and tell us your variables and how many groups you are comparing. We will tell you which test the rubric actually wants and what the justification sentence should say.

Related reading: IB Biology IA Data Collection Checklist · Our IB Biology programme