meta analysis
Last reviewed 01/2018
This has proved a very useful technique, especially in areas where a large amount of data has been collected apparently with an inconclusive result. Published literature is surveyed and results from all well conducted trials (blinded, adequate clinical details, clear drop-out criteria etc.) are pooled and the data re-analysed.
Because it is usually only positive results which are published there is a tendency for type I errors to be included (wrongly rejecting the known hypothesis - assuming a difference where none exists). It may be useful therefore to include data which is equally soundly based but has not been published. Such data may be subject to type II errors (wrongly accepting the null hypothesis - assuming no difference when actually there is a difference). These errors generally arise because the number of subjects is to small, but combining them in a meta analysis reveals the correct outcome.