Distinction Between Two Statistical Terms: Multivariable and Multivariate Logistic Regression

Dear Editor, Two statistical terms, multivariate and multivariable, are repeatedly and interchangeably used in the literature, when in fact they stand for two distinct methodological approaches.1 While the multivariable model is used for the analysis with one outcome (dependent) and multiple independent (a.k.a., predictor or explanatory) variables,2,3 multivariate is used for the analysis with more than 1 outcomes (eg, repeated measures) and multiple independent variables.1 However, the terms are sometimes used interchangeably in the literature as not many researchers are attentive to the distinction. The difference between these two terms was brought to attention by Hidalgo and Goodman in 2013.1 Yet, some researchers continue to report these terms interchangeably. For instance, in a recent article published in Nicotine and Tobacco Research,4 although the data analysis approach was detailed, they used the term “multivariate logistic regression” models while their analysis was based on “multivariable logistic regression”; this was emphasized in Table 2’s legend in the same article. This has also occurred in other articles published at Nicotine and Tobacco Research5,6 and elsewhere.7–9 Therefore, we use this opportunity to highlight a clear distinction and identify the nuances that make these types of analyses different from one another.