Oh, shoot! I forgot about that! Estimating the influence of uncontrolled variables

In many experimental and applied contexts, behavior analysts collect data on a variety of variables that have a logical relationship to a behavior they want to describe, predict, or control. Though experimental control of independent variables is ideal, it is not always possible, especially in applied contexts. However, this doesn’t mean we have to throw our hands up and claim ignorance. Instead, we can use modeling techniques already prevalent in behavior analysis to estimate the precise direction and size of influence that each variable has on behavior. This chapter reviews (1) how modeling fits into the larger conversation on the description, prediction, and control of behavior; (2) basic types of regression and classification models behavior analysts are likely to find useful; (3) how to assess whether a model is doing a good or poor job describing our data by way of loss metrics; and (4) how to interpret the free parameters of models that describe our data well.