Run a t-test on two groups, paired scores or one sample T-TEST
Paste the scores, or type the n, mean and standard deviation from a paper. You get t, the degrees of freedom, p both ways, the difference with its confidence interval, Cohen's d, and the sentence for a write-up.
The test
The numbers
One value a line, or separated by spaces or commas. Paste a column straight from a spreadsheet.
The result
For the write-up
How it works
T-TEST works from two columns pasted from a spreadsheet, or from the summary figures printed in a paper. Levene's test sits beside the independent test so the choice between Student and Welch is in view, and every figure is held to SciPy. Nothing typed here leaves this tab.
What it does, and what it does not
A t-test asks whether a difference in means is larger than the scatter in the data would produce by chance. This page runs the four common versions: two independent groups with equal variances assumed (Student) or not (Welch), the same people measured twice (paired), and one group against a fixed value. The p-value is worked out from Student's t distribution directly, not looked up in a table, and the confidence interval uses the same distribution, so the two always agree.
It does not test three or more groups (that is an analysis of variance), and it does not run the rank-based tests, Mann–Whitney and Wilcoxon, that suit small samples far from normal. It does not check normality beyond what you can see, and it does not work out how many people a study needs. It does not correct for running many tests. Missing values are simply left out: in paired data a line with a value missing on either side is dropped, and the count says so.
The convention used for each figure is stated beside it. Cohen's d for two groups is over the pooled standard deviation whichever test is chosen, and Hedges' g is shown with it for small samples. The one-tailed p is in the direction the difference went, which is only fair if that direction was predicted. Every figure is held to SciPy in this site's own tests. To describe a single column first, FIGURES does that, and CHI-SQUARE tests counts rather than scores.
Doing this often? OBSCURA Research runs t-tests on the columns of a whole data set, with ANOVA, the non-parametric tests and regression beside them, and APA tables saved to Word. On its own, US$2 a month; with everything else, US$5 a month or US$48 a year. 7 days free.