Meta-analysis

A meta-analysis isstatistical procedurecombiningnumberexisting studies. The first meta-analysis was performed by Karl Pearson1904,an attemptovercomeproblemreduced statistical powerstudiessmall sample sizes; analyzingresults fromgroupstudies can allow more accurate estimationeffects.

Although meta-analysiswidely usedmedicine today,meta-analysis ofmedical treatment was not published till 1955. In1970s more sophisticated analytical techniques were introducededucational research, starting withworkEugene V. Glass.

Becauseresults from different studies investigating different dependent variablesmeasured on different scales,dependent variable inmeta-analysissome standard measureeffect size, such asstandard score equivalent todifference between means (d), or an odds ratio.

A weakness ofmethodthat sourcesbiasnot controlled bymethod. A good meta-analysisbadly designed studies will still resultbad statistics.

See also:

• Simpson's paradox
• Selection bias

Outspoken critics:

• Ray Hyman