In a 2005 paper in PLoS Medicine “Why Most Published Research Findings are False,” Dr. Ioannidis provided convincing quantitative simulation evidence that most scientific research findings are not replicable and are therefore false. Discuss two factors he reported that are of particular interest to experimental design in terms of establishing causation.

Question

What will be an ideal response?

Answer 1

Ans: Varies.

There are two key factors Ioannidis identified in false research findings that are particularly interesting about experimental design. First, studies that use small samples are more likely to report findings that are not replicated and end up as false. These findings, even though they achieved statistical significance, may be due to inadvertent sample bias--the flukiness of small numbers. Dr. Ioannidis also noted that studies that show small effect sizes are difficult to replicate and these findings too are more likely to be false. The effect size refers to the strength of the predicted or hypothesized relationship between the independent variable and dependent variable. The first question to ask when reading research is whether the finding is replicable, or more precisely, has the finding been replicated by an independent group of researchers. The second question is how robust is the effect.