“To what degree do hypotheses confirmed by correlational human studies are later vindicated by experimental studies?”
That depends on what you mean by a ‘correlational study.’ People who analyze observational data with causality in mind spend a lot of time thinking about potential confounding and what to do about it. For example, here’s an analysis of a very large longitudinal dataset with the aim of determining a causal effect:
which does very sensible things. If you look at sensible analyses of observational data, then ‘the vindication rate’ will be related to how often the needed assumptions actually hold. If you look at non-sensible analyses (e.g. that aren’t adjusting for confounder bias, and so on), then it’s just garbage, no reason to expect better than chance then.
That depends on what you mean by a ‘correlational study.’ People who analyze observational data with causality in mind spend a lot of time thinking about potential confounding and what to do about it. For example, here’s an analysis of a very large longitudinal dataset with the aim of determining a causal effect:
http://www.hsph.harvard.edu/wp-content/uploads/sites/1138/2012/09/ije_2009.pdf
which does very sensible things. If you look at sensible analyses of observational data, then ‘the vindication rate’ will be related to how often the needed assumptions actually hold. If you look at non-sensible analyses (e.g. that aren’t adjusting for confounder bias, and so on), then it’s just garbage, no reason to expect better than chance then.