SobelTest

SobelTest refers to a statistical test used to assess the significance of indirect effects in mediation analysis. Mediation analysis aims to understand how an independent variable (X) influences a dependent variable (Y) through a third variable, the mediator (M). The indirect effect is the product of the path coefficient from X to M (a) and the path coefficient from M to Y (b), so it is represented as a b.

The SobelTest was developed by David Sobel and is one of the earliest and most widely used methods for testing the significance of this indirect effect. It approximates the sampling distribution of the indirect effect using a normal distribution. The test calculates a z-score by dividing the estimated indirect effect (a b) by its standard error. The standard error is calculated using the delta method, which involves the variances and covariances of the individual path coefficients.

A p-value is then derived from this z-score, allowing researchers to determine if the indirect effect is statistically significant. If the p-value is below a predetermined alpha level (commonly 0.05), the indirect effect is considered significant, suggesting that the mediator plays a role in the relationship between the independent and dependent variables.

While historically significant, the SobelTest has limitations. It relies on assumptions of normality and can be less powerful than alternative methods, especially with smaller sample sizes. More modern approaches, such as bootstrapping, are often preferred for their greater accuracy and fewer assumptions. However, the SobelTest remains relevant for understanding the development of mediation analysis and is still encountered in older research.