Researchers often want to aggregate and synthesize evidence from multiple studies of the same question to get a more precise estimate of the effect of a particular intervention.
statistics
Description
Problem 1
Researchers often want to aggregate and synthesize evidence from multiple studies of the same question to get a more precise estimate of the effect of a particular intervention. These sorts of studies are called meta-analyses and are very common in medical and psychological research where many experiments of varying size are often run in many different contexts. We are going to replicate a meta-analysis using an example dataset from the meta suite in Stata of a series of experiments analyzing the effect of teacher expectations on student performance (pupiliq.dta). From the description of the data available here (pp. 19) This example describes a well-known study of Rosenthal and Jacobson (1968) that found the so-called Pygmalion effect, in which expectations of teachers affected outcomes of their students. A group of students was tested and then divided randomly into experimentals and controls. The division may have been random, but the teachers were told that the students identified as experimentals were likely to show dramatic intellectual growth. A few months later, a test was administered again to the entire group of students. The experimentals outperformed the controls. Subsequent researchers attempted to replicate the results, but many did not find the hypothesized effect. Raudenbush (1984) did a meta-analysis of 19 studies and hypothesized that the Pygmalion effect might be mitigated by how long the teachers had worked with the students before being told about the nonexistent higher expectations for the randomly selected subsample of students. We will be working with the Raudenbush data set in R. First, load the data via the function read_dta from the haven package (part of the broader tidyverse).
