Time-Series Design

Anecdote: took ~17 years for a new method to enter clinical practice.

Pre-Post Design with Concurrent Controls. Very common. There is no second group.

Sample -> Treat -> Measure -> No Treat -> Measure -> ...

The good thing about this is that “innate characteristics are eliminated as confounders”. But it sucks because there may be regression to the mean. Learning effects. You would not use this for drug efficacy!

Post-Test only with Controls. TODO: Complete.

Removed Treatment Design. A single group acts as its own control. Over equal periods of time, you add and remove the intervention.

Double Pretest Pre-Post Design. Same as Removed Treatment but you Measure… wait… Measure again… and then introduce intervention… and then measure as you would. Measurements must be equally spaced! Lets you know if somthing funky’s going on.

Pre-Post Design. Very common, but this is the weakest (because regression towards the mean).

TODO: Relative Authority table. Start with RCT… Crossover… Pre-post… Post-test-only.