Skip to main content

Validity

General soundness and credibility of findings and conclusions.

There’s study Validity and test Validity. The latter has to do with instrumentation.

Study Designs​

Observational Study Designs​

  • Prospective: select on exposure, follow → time
  • Retrospective: select on exposure, follow < - time
  • Cross-Sectional Study: no time. slice. No incidence (same reason)
  • Case-Control Study: always Retrospective.
  • Crossover Study: Cases serve as their own controls (think seasonal allergies) - reduce random error and confounding.

Sidenote: Reconstructing temporality from clinical notes. Harder than it sounds. George worked on this.

This stuff’s on the exam btw.

Experimental Study Designs​

Discussion on creating control groups. Pre-test, post-test. Tradeoffs.

Randomization really only works with really large numbers. With a small number you may not be able to remove all confounders. And removing them is the entire point of the exercise.

Complete Factorial Design: assess the impact of multiple interventions (permutations) without creating arms for each in an RCT.