Effect Sizes for Continuous Data

Meta-analysis commonly encounters continuous outcomes when researchers examine variables measured on a numerical scale, such as blood pressure, body weight, temperature, or laboratory measurements. This session introduces the fundamental effect measures used to synthesize continuous data, particularly the mean difference (MD) and standardized mean difference (SMD). Use the mean difference when studies measure the same outcome on the same scale, allowing results to be expressed in the original units. In contrast, the standardized mean difference is useful when studies assess the same underlying construct using different measurement scales. Students will explore Cohen’s (d) and Hedges’ (g), two commonly used standardized effect measures, and learn how to distinguish standardized from unstandardized effects when selecting an appropriate method for meta-analysis.

The second component focuses on calculating effect sizes and quantifying their uncertainty using summary statistics, including group means, standard deviations, and sample sizes. Students will learn the basic principles of estimating the variance and standard error of MD and SMD, which are essential for evaluating the precision of individual study estimates and combining results across studies. The course will focus on Hedges’ (g), which corrects Cohen’s (d) to reduce small-sample bias in the standardized effect estimate. Students will also examine how to interpret an effect's direction and magnitude, considering which group serves as the reference and how the outcome is measured. These concepts provide the statistical foundation for understanding how continuous outcome data contribute to pooled effect estimates in meta-analysis.

The final component addresses practical challenges associated with continuous outcome data, particularly studies using before-and-after measurements and paired designs. Unlike independent-group studies, paired designs measure the same individuals or matched units, so correlation between measurements can affect the variance and standard error of the estimated effect. Students will learn why the appropriate effect-size calculation depends on the study design and the information reported in the original research. They will also recognize the importance of identifying whether summary statistics represent independent groups, repeated measurements, or change scores before conducting statistical synthesis. By the end of this session, students should be able to distinguish MD from SMD, explain the differences between Cohen’s (d) and Hedges’ (g), calculate or identify the required summary statistics, and interpret effect sizes and their uncertainty appropriately in a meta-analysis.