Multi-collinearity in Research and Wayforward

Abstract

In multi-linear regression, there will be two or more independent variables also known as predictor variables. The main task in multi-linear regression is to find how the predictor variable impacts the dependent variable when there is a change in predictor variables by one unit. However, without testing multi-co linearity between these predictor variables, such a model will create difficulties in terms of defining the real impact of the predictor variable (independent) on the predicted variable (dependent). This situation makes us aware in terms of variable selection while conducting multi-linear regression to find the exact impact. To explain this issue, 14 students were selected and asked to fill up the form with their marks obtained in the recent examination, the number of study hours at home, the number of hours hanging out with friends, and the number of copies added during an exam. The research aimed to observe if there is a multi-co linearity issue between predictor variables (number of study hours, number of hours hanging out with friends, and number of copies added during an exam). SPSS-version 25 is used to find a correlation matrix between predictor variables, matrix scatter plot, the value of Tolerance, Variance Inflation Factor (VIF) value, condition index, and variation proportion. Multi-co linearity issue observed in two predictor variables thus further solution explained.