National Population Studies & Comprehensive Management Institute SAMPLE SIZE IMPACTS ON HIGH LEVERAGE COLLINEARITY - ENHANCING OBSERVATIONS

the latest known source of multicollinearity, a nonorthogonality of two or more explanatory variables in multiple regression models, is high leverage points. Interpreting a fitted regression model may become impossible by the influential impacts of multicollinearity. In this paper, we attempt to investigate the impact of different sample sizes as one of the main causing factors of high leverage points to be collinearity-influential observations in non-collinear data. To do so, the influence of changing sample size on High Leverage Collinearity-Influential Measure (HLCIM) and Condition Number (CN) was studied. According to the simulation results, by increasing the percentage of high leverage points for each magnitude of contamination and fixed sample size and also by increasing magnitude of contamination for each percentage of high leverage point and fixed sample size, the CN and the absolute HLCIM values increase. The simulation results have been confirmed by a well-known real data set.