|
How To... > Work with Statistical Objects > Test Hypotheses > Test Correlation
Conceptually, there is a difference between testing whether a least-squares slope is different from zero and testing whether a correlation coefficient is different from zero. The first assumes that a response attribute’s values are being predicted from a predictor’s values. But in a test for correlation, the two attributes play a completely symmetric role.
| 1. | Create a hypothesis test object by dragging one from the or by choosing . |
| 2. | Choose from the test’s pop-up menu. |
| 3. | Drop numeric attributes on the prompt at the top of the test. |
Because a test of correlation only requires two numbers, the count and the correlation coefficient, you can enter these numbers directly in the test without needing raw data.
In the example, the same “play” data from the description of Test Slope is used to illustrate Test Correlation. If you construct these two tests side-by-side using the same data, you’ll see that student’s t is the same for both. This means that if the slope is significant, so is the correlation coefficient.
| 
