Counts of cases, never percentages. Chi-square asks whether two categorical variables are related.
The independent-samples t test asks whether two groups differ on one numeric measure. Exactly two groups.
Use this to check a result a classmate, a textbook, or a published article reports, when you have the summary statistics but not the data.
Results
What this page will and will not tell you
- It recomputes the test and the effect size, so a transcription slip between your output and your write-up shows up here.
- It flags the assumption problems that actually sink these two tests: expected counts under 5, a table too small for the approximation, variances far apart, a group too small to lean on normality.
- It reports the effect size next to p, because a significant trivial difference and a non-significant large one are both easy to write up wrongly.
- It cannot tell you whether the test is the right test, whether your cases are independent, or whether the variable means what you say it means. Those come from the design, and they are what the codebook and the prospectus were for.
- The t test shown is Welch’s, which does not assume equal variances and is the better default. Say in your write-up that you used it.
Serves the analysis stage in MC 451 and MC 501, and Module 6 in MC 427.