### t-tests

When reporting a *t*-test you should include the following information.

**State whether you used an independent or paired**You should use a paired*t*-test:*t*-test if you took repeated measurements of the same variable from the same sample. Otherwise use an independent*t-*test.**Check and state that your data is normally distributed:**Your data should follow a normal distribution. Check this by plotting the data on a histogram or scatter graph. For an independent*t-*test you should plot the data for each group. For a paired*t*-test you should plot the paired differences. If the chart peaks in the middle and is roughly symmetrical then your data is normal. Otherwise, you should use a Wilcoxon or Mann-Whitney test.**Check and state that variances are equal (for independent**you can use Levene's test for this. If variances aren't equal, you can correct for this by adjusting the degrees of freedom using the Welch-Satterthwaite method and by not using the pooled estimate for the error term for the t-statistic. Some statistical softwares will do this for you automatically - look for a row in the output called something like "Equal variances not assumed". Alternatively, you could use a MannWhitney test.*t*-tests):**State the degrees of freedom and the test statistic:**Write this as*t*(df) =*t*-statistic.**Include a measure of precision:**generally a 95% confidence interval.

If you want to plot your results you should use a box plot (for exploratory research) or a confidence interval plot (if you are testing a hypothesis).

See the SAMPL guidelines for more general information about how to report statistics, and this guide to learn more about common statistical tests.