Chowdhury's Non Parametric Mediation Model

Chowdhury's Non Parametric Mediation Model

The non-parametric mediation statistics is a non-parametric regression based statistics to predict the influence of a qualitative independent variable when another independent variable have an effect on a qualitative dependent. For example variable A has indirect effect on variable B but the also the variable C has a direct effect on variable B; then, we can claim the variable C as a mediator and it might have a mediation effect on variable C. It establish the mediation effect of variable C we need to check whether the strength of effect of variable A on variable B goes down in presence of variable C in independent list, when variable A and variable C are corelated to each other. Non-parametric correlation coefficients such as Kendal's Tau, Spearman's correlation, etc. could be computed to check relationships between 3 or more variables when the study was conducted with related sample design. Then we apply the non-parametric regression models such as Logistic regression model to have better resolutions in data in the mediation model. In mediation model the effect of mediator on dependent variable is generally higher than other associated independent variables. The relationships between other independent variable on dependent variable tens to zero in presence of mediator (independent) variable. A sample non-parametric model has been presented  in the following image:

Fig. 1 Mediation effect of Work Shift Duration on Symptoms of Musculoskeletal Ailments (Adopted from Sanjog, Patel, Chowdhury and Karmakar).

According to the model III of Fig. 1, variables WD,  WSD and SMA are corelated to each other, however, in present of WSD the correlation or regression coefficient value between WD and SMA is significantly decreased in presence of WSD. Hence, WSD (mediator) is having the mediation effect in relationship between WD and SMA. Therefore, WSD is more valuable predictor of SMA than the WD of workers in the context of plastic chair manufacturing.