Canal
What Are We Modeling For? A Case for a Holistic Analytic Approach (303770)
*Vasile Alexandru Suchar, Whitman College
Keywords: Regression; SEM; ML
In recent years it had been more frequently argued that prediction should become the fundamental aspect of the scientific process, which would result in (1) more mature hypotheses being generated, (2) modeling toolset prioritizing prediction, and (3) ultimately resulting in a better understanding. The ultimate functionality of sciences is to understand and predict the consequences of actions or changes in systems. Under this premise, I will argue that sciences should not focus only on prediction, but they should welcome holistic analytic approaches combining both explanatory and predictive statistical inference methods. The argument is not in the favor of indiscriminate use of statistical methods, or either statistical method. But, scientists must be “statistically fluent”, and capable to choose the appropriate methods. In this presentation, I will show the differences between prediction and explanation modeling, present the analytic toolset at our disposal and make a case for multiple modeling approaches. By using several examples, I will show how a holistic analytic approach can be used for predictors’ and model’ “triangulation” and guide outgoing research.