Descriptive estimands, causal estimands, and avoiding the jungle of adjusted associations ‘in between

The first recommendation for quantitative studies in JECH’s recently published methodological recommendations1 for prospective authors reads:

Be upfront about the research study’s intention (this should link directly to the aim) – is it to describe, predict, or estimate causal effects (these are all very different tasks).1

Yet the Journal regularly receives manuscripts which do not clearly articulate their purpose. Indeed, even our own studies will no doubt sometimes have blurred these three distinct research aims—description, prediction, or estimation of causal effects.1–3 While the distinction between predictive and causal research aims has been explicitly discussed before by Hernán et al,4 there remains an ill-defined space between descriptive and causal research. This is the space we will explore in this article.

Causal and descriptive estimands: the ‘What If’ and the ‘What Is’

Estimands represent the quantities of interest that a study sets out…

Read the full article ›