Type: statistical procedure

Measurement level: Correlates: at least one nominal and at least one metric, Happiness: metric.

Just as in an ANOVA, in an ANCOVA the total happiness variability, expressed as the sum of squares, is partitioned into several parts, each of which is assigned to a source of variability. At least two of those sources are the variability of the correlates, in case there is one for each correlate, and always one other is the residual variability, which includes all unspecified influences on the happiness variable. Each sum of squares has its own number of degrees of freedom (df), which sum up to Ne -1 for the total variability. If a sum of squares (SS) is divided by its own number of df, a mean square (MS) is obtained. The ratio of two correctly selected mean squares has an F-distribution under the hypothesis that the corresponding association has a zero-value.

In an Analysis of Covariance, the treatment means for all levels of the nominal correlate are 'adjusted' for differences in the mean values of the metric correlate.