## The intercept in regression

The *constant* term in a regression ANOVA table tests whether the mean of the outcome measure is zero. It is usually not important if the significance test result in a positive outcome, but can be useful if the outcome is a difference.

For example suppose **digit span forward** and the **GOAT** tests are both measured at the same two times on the same person. We wish to see if the difference over time in **digit span forward** is explainable by the difference over time in the **GOAT**. This can be done by taking the difference between the two time points for each test score. Then we need to regress the **GOAT difference** on **digit span forward difference**.

In the output we now check the t-test for the constant (or intercept) term in the regression. Seeing if the constant term is zero tells us if there is a non-zero difference in digit span forward scores, averaged over individuals, even after accounting for their difference in GOAT scores.

See also: **Everitt B, and Hay D.** (1992) p.90-91 *Talking About Statistics: A Psychologistâ€™s Guide to Design and Analysis*. Edward Arnold. This text looks at the intercept to see if the difference in dysphoria scores is the same across group (a group x dysphoria (covariate) interaction).