Proposed Measure for Effect Size in Mediation Analysis with Solution to Some Mediation Process Problems

Abstract

In this paper, the effect size measures was discussed, which are useful in many estimation processes for direct effect and its relation with indirect and total effects. In addition, an algorithm to calculate the suggested measure of effect size was suggested that represent the ratio of direct effect to the effect of the estimated parameter using the Regression equation of the dependent variable on the mediator variable without using the independent variable in the model. Where this an algorithm clear the possibility to use this regression equation in Mediation Analysis, where usually used the Mediator and independent variable together when the dependent variable regresses on them. Also this an algorithm to show how effect of the parameter estimate of the equation Regression that contain the mediator variable only on the direct effect. The Robust Ridge Regression approach that suggested by (Askin and Montgomery,1980) is used also in mediation analysis as suggested procedure by used it in mediation analysis to overcome on the linear dependency  in Mediation Analysis, furthermore we remedy the outlier points problem in the data. All that was explained through generated and real data.