|
The paired t test assumes that you have sampled your pairs of values from a population of pairs where the difference between pairs follows a Gaussian distribution. If you want to test this assumption with a normality test, you need to go through some extra steps:
| 1. | From your data table, click Analyze and choose "Remove baseline...". |
| 2. | On the Remove Baseline dialog, define the baseline to be column B, and that you want to compute the difference. |
| 3. | View the results table showing the differences. Click Analyze and choose Column statistics. Note that you are chaining two analyses, first subtracting a baseline and then performing column statistics on the results. |
| 4. | Choose the normality test(s) you want. We recommend D'Agostino's test. Note that none of the normality tests are selected by default, so you need to select one. |
| 5. | If the P value for the normality test is low, you have evidence that your pairs were not sampled from a population where the differences follow a Gaussian distribution. Read more about interpreting normality tests. |
If your data fail the normality test, you have two options. One option is to transform the values (perhaps to logs or reciprocals) to make the distributions of differences follow a Gaussian distribution. Another choice is to use the Wilcoxon matched pairs nonparametric test instead of the t test.
|
