1. Enter your data
Paste your raw trial data, type it in by hand, or load one of the example datasets to explore. The first column is your independent variable (group name or IV value). The remaining columns are individual trial measurements.
Data table
Click any cell to edit. Each row is one IV value or treatment group. Each column after the first is a single trial.
Or load an example
2. Descriptive statistics & outlier detection
A summary of each group's measurements. Outliers (values more than 1.5ΓIQR from the nearest quartile) are flagged in red. Box-and-whisker plots help you see the distribution at a glance.
Summary statistics
Box-and-whisker plots
3. Assumption checks IB DP
Most parametric tests (t-test, ANOVA, Pearson's r) assume your data is approximately normally distributed and that groups have similar variances. If these are violated, a non-parametric alternative is better. This step runs those checks automatically.
4. Which statistical test?
Use the decision tree to trace a path to the right test. The tool will pre-select branches based on your data and assumption checks, but you can override them.
5. Test result
Test statistics
Effect size
Step-by-step working
6. IA results paragraph builder
A structured results section you can adapt for your IA. It includes descriptive statistics, assumption checks, test choice, inferential result, effect size, interpretation and cautions. Do not paste it unchanged. Edit it so it matches your research question, graph, uncertainty and biology.
Guide & reference
Vocabulary and quick reference for every test in the tool. Use this alongside the workflow if you get stuck on terminology.
Which test to use - quick reference
Independent t-test (pooled)
Compares the means of two independent groups on a continuous measurement (e.g. control vs treatment). Assumes normality and equal variance. Reports t, df, p and Cohen's d.
Welch's t-test
The same purpose as a pooled t-test (compare two means) but does not assume equal variance. Use this when Levene's test flags unequal variance. Uses the Welch-Satterthwaite df approximation. Modern statistics references often recommend Welch's as the default over the pooled version.
Paired t-test
Compares two measurements taken from the same subjects (e.g. heart rate before and after exercise). Tests whether the mean of the differences is zero.
One-way ANOVA
Compares means across three or more groups. If significant, Tukey HSD finds which specific pairs differ. Effect size: eta-squared (Ξ·Β²).
Chi-squared (ΟΒ²)
For count / categorical data. Goodness of fit tests whether observed counts match expected proportions β enter your ratio in Step 5 as e.g. 1:1, 3:1 (Mendelian monohybrid), 9:3:3:1 (dihybrid) or any other proportion. Test of independence checks whether two categorical variables are associated. Reports ΟΒ², df and p.
Pearson's r
Measures the strength and direction of a linear relationship between two continuous variables. Use only when data is linear and normally distributed. Reports r (-1 to +1) and p.
Spearman's rank (Ο)
Non-parametric correlation. Use when data is ordinal, non-linear (but monotonic), or non-normal. Based on ranks, not raw values.
Mann-Whitney U
Non-parametric alternative to the independent t-test. Use when normality is violated. Tests whether one group tends to have higher ranks than the other.
Kruskal-Wallis H
Non-parametric alternative to one-way ANOVA for 3+ groups. Use when normality is violated. Tests for differences in rank distributions.
Key vocabulary
p-value
The probability of observing your data (or more extreme) if the null hypothesis were true. Lower = stronger evidence against the null. Conventional threshold: p < 0.05.
Null hypothesis (Hβ)
The default position: "there is no effect / no difference / no relationship." Tests assess evidence against this position, never for the alternative.
Degrees of freedom (df)
Number of independent values free to vary. Depends on the test and sample size. Always reported alongside the test statistic.
Effect size
How big the effect is, independent of sample size. Cohen's d for t-tests (0.2 small, 0.5 medium, 0.8 large). Ξ·Β² for ANOVA (0.01 / 0.06 / 0.14). Required by modern IA standards.
Outlier
A value far outside the rest. Tukey's rule: more than 1.5ΓIQR below Q1 or above Q3. Always justify removal - never delete data silently.
Standard error (SE)
How precisely your sample mean estimates the population mean. SE = SD / βn. Use SE for error bars, not SD, when comparing means.
Normality
Whether your data follows a bell curve. Tested with Shapiro-Wilk. Required for parametric tests (t, ANOVA, Pearson). If violated, switch to non-parametric.
Equal variance
Whether groups have similar spread. Tested with Levene's test. Required for the standard t-test and ANOVA. Welch's correction handles unequal variance.
Critical reminders for your IA
· Report descriptive statistics first (mean, SD, n), test result second.
· If you check assumptions, say so - and explain what you did if they were violated.
· Always quote an effect size alongside the p-value. A small p with a tiny effect is not biologically meaningful.
· Link the conclusion back to your hypothesis and the underlying biology.
About the calculations
All tests are computed in your browser - no data leaves your device. Distribution functions (t, F, ΟΒ²) use standard numerical approximations (incomplete beta and incomplete gamma via continued fractions). The Shapiro-Wilk implementation uses a Royston-style approximation. For borderline results, unusual datasets, or final IA submission, students should verify the result in Jamovi, R, SPSS or another statistics package.