Skip to main content

Study Design

All study designs have a well-defined population PP from which groups of subjects studied and outcomes OO that are measured. The good definition is about reproducibility and replicability. What’s the difference? Same data, same methods, same results is the former. New data, same methods, consistent results is the latter.

You have descriptive and analytic studies. The latter involves at least two exposure states.

A big thing about causality is temporality. Exposure must precede outcome.

Research Question drives the study design, which determines the parameters you legitimately estimate.

Descriptive:​

Cross sectional study measures outcome and exposure at the same time. It is just a snapshot. You are not establishing exposure → outcome. You are not saying anything about causality. Think of how timelines (e)------(o) are smushed into an Excel spreadsheet. You just get (e)s and (o)s, the lines vanish. No temporality, no causality. Now you can do prevalence but you cannot do incidence! And really, you can mostly only estimate prevalence. The other is measures of association (think contingency table!)

Population ---- Controls ----- Exposed
                         |---- Not Exposed

Odds increase very quickly as the outcome becomes more common.

TODO​

  • Why not incidence with x-sectional study?