|Though there are statistical methods that handle relationships between many variables, very often we look at whether one variable, the dependent variable, is systematically associated with, correlated with, another variable and typically this is framed as the relationship between a predictor variable and a dependent variable: does gender predict early opting out in therapy or not? Here early opting out is the dependent variable and gender the predictor variable.|
Oddly enough, statisticians refer to this as “univariate statistics” despite the fact that there are two variables involved (just like the rather bizarre terminology “individual therapy” for therapies which generally involve both a client and a therapist: in both cases perhaps it’s only the perceived “dependent” that gets counted!) They reserve “multivariate” statistics for the situation in which you have more than one dependent variable. For example: if we have both baseline PHQ-9 (depression) and GAD-7 (anxiety) scores, is there some relationship in which gender, say, predicts the two. Equally rather illogically, having more than one predictor variable but only one dependent variable is still univariate statistics: for example, is there a relationship between gender and employment status and early opting out of therapy? Where you have more than one predictor variable you have the possibility of an “interaction”: for example both being male and being unemployed might be associated with early opting out, say each doubles the likelihood of early opt out. However, the relationship might be more than additive and being both male and unemployed might raise the likelihood of early termination eight times.
Try also … #