Two-Sample Comparisons
To compare the means of two samples, the t-test is usually employed unless the data are ranks, or are so far from normally distributed that you are nervous about the validity of the analysis. To do a t-test, you can use an on-line tool that allows you to enter
- either your raw data or descriptive stats calculated from those data here
- just the descriptive stats here or here
- just the raw data here or here or here
Note that there are lots of options for doing a t-test, depending upon whether you have equal variances or paired data, and whether you want a 1-tailed test or a 2-tailed test. You can learn more about these options here. When in doubt, assume equal variances and unpaired data, and calculate the 2-tailed test.
A worked example
Using the example at the bottom of the confidence limits page, we can compare the weights of male and female ptarmigan using a t-test. For the first sample of 19 birds, the results from the t-test are t = 1.18, P = 0.25, n = 10 males, 9 females. From this analysis you would report: “males weighed 14 g, on average, more than females but the difference is not statistically significant (t = 1.18, P = 0.25, n = 10 males, 9 females)”.
Note that sometimes you will get a statistically significant result even when the 95%CLs overlap, but if they do not overlap, then the difference is always statistically significant at P≤0.05.