AI Statistical Test Advisor

The right test follows from three facts about your study: what your hypothesis compares or relates, how each variable is measured, and how the observations were collected. Describe all three precisely — for instance, 'compare job satisfaction (a 5-item Likert composite) across three management levels, independent groups, n = 240' — and the advice will be specific enough to act on. Vague inputs such as 'analyse my survey' can only produce generic suggestions.

Measurement level matters most. A single Likert item is ordinal and is usually analysed with non-parametric tests, whereas a composite score averaged across several items is commonly treated as continuous. Group comparisons call for t-tests or ANOVA and their non-parametric equivalents; relationships call for correlation or regression; and latent constructs measured by multiple items, with several hypothesised paths between them, usually point to structural equation modelling in AMOS or SmartPLS.

Confirm assumptions before you trust any result: normality, homogeneity of variance, independence of observations and, for regression, multicollinearity and outliers. Then report the full picture — the test statistic, degrees of freedom, exact p-value and an effect size — rather than significance alone. If your supervisor prefers a particular approach, discuss the recommendation with them; the tool explains the reasoning so that you can defend the choice in your viva.