How does the nature of my outcome variable influence my choice of which type of multivariable analysis to do?

As shown in Table 3.1, the choice of multivariable analysis depends primarily on the type of outcome variable that you have.

What type of multivariable analysis should I use with an interval outcome?

With an interval variable (also called continuous) each unit (interval) of change on the scale has an equal (numerically) quantifiable value. Examples of interval variables are blood pressure, body weight, and temperature. In these examples, a one-unit change at any point on the scale is equal to a millimeter of mercury, a pound (or kilogram), or a degree, respectively.

Interval outcomes can be analyzed with multiple linear regression or with analysis of variance. Either technique will yield the same answer, assuming you set up the models in similar ways. In general, multiple linear regression is more commonly used with observational data, and analysis of variance is more commonly used with experimental designs. Multiple linear regression is more commonly used in the medical literature, and analysis of variance is more commonly used in the behavioral literature.

Because multiple linear regression is easier to explain, easier to set up correctly, and easier to interpret than analysis of variance, I have focused on this technique in this text.