Validity

"At best we can only know what has not yet been shown to be false"
(Cook & Campbell, 1979: presenting a Popperian view of science)
 
Four key questions in science:
1) Is there a relationship?
2) Is it causal?
3) If causal, what are the constructs involved
4) How generalizable is the relationship between (among) constructs

Construct validity
Do variables accurately reflect what they are intended to?
Can depend upon:
· operational definitions of hypothesized causal/intervening variables
· use of standardized protocols (to minimize error)
Internal validity (Causality)
Depends upon:
1) Statistical conclusion validity
· Is there sufficient covariation to infer cause?
2) Internal validity related to experimental design & methods
· Are there any confounds?
External validity (generalizability)
Depends upon:
· The population selected for study
· Representative sampling of that population
The best evidence for external validity comes from replication and converging evidence.

Often there are trade-offs among the different forms of validity with a priority on internal validity in most science (except for applied research, which emphasizes external validity and/or ecological validity).