Causal Language in Structural Equation Modeling

Abstract

Structural equation models (SEM) are useful tools for proposing theoretical relationships, causal or not, between multiple variables. However, they do not allow us to confirm the existence of cause-effect relationships in the absence of an experimental design. We evaluated the occurrence of inadequate expressions of causality in non-experimental articles, published in Spanish, using SEM as data analysis. After a systematic review of studies included in five databases, we reviewed the 188 selected articles. Of these, 63 studies (33. 5%) used correct language in the title and abstract; the remaining 125 (66. 5%) used biased or incorrect language in at least one of the sections. These inappropriate interpretations may lead the reader to erroneous conclusions, threatening the rigor of scientific research. Any causal conclusions derived from SEM should be formulated as associations between variables, warning of the nonexperimental nature of the study, and suggesting alternative explanations.