I was looking through some of my old notes and came across this list of “data analysis principles”.
- answer the question
- understand where the data came from
- make a graph
- does the result make sense?
- if it seems too good to be true, it probably is
- form a diagnostic checklist: think of ways things could go wrong and how they might be revealed
- follow up on all aberrations
- every estimate needs an SE
- consider taking logs
- consider taking differences
- use common axis scales
- avoid normality assumptions
- remember Simpson’s paradox
I also found this list of “key statistics concepts”.
- separate the population and the sample, the parameter and the estimate
- imagine doing it again: estimates have distributions
- suppose there were no effect; are the observed data unusual?