INVESTIGATION OF PANEL MODELLING TECHNIQUES IN THE PRESENCE OF COLLINEARITY REGRESSORS

Abstract

In order to estimate the presence of collinearity, this study used the fixed effect model, pooled effect model, and random effect model in the presence of collinearity regressors. In respond one, data were simulated under several types of collinearity verifiable at varying sample sizes (-0.1, -0.5, -0.9, 0.1, 0.5, and 0.9) in the Monte Carlo simulation. For the models estimators, two regressor estimators were used. Simulations were run at various panel structures and collinearity regressors in the Monte Carlo study. The trial was conducted ten thousand times (10000), and the accuracy of the model estimation was assessed using the Root Mean Square Error (RMSE). The results of the study showed that the estimation of the small sample panel structure model. While following time series lengths (5, 10, 10, 30, 60, and 60) have 10,000 repetitions of the experiment conducted in the R environment. The Root Mean Square Error (RMSE) was used to assess the models. The RMSE values for the fixed and random model are fluctuated as the collinearity levels grew in all of the scenarios that were taken into consideration. Based on the analysis, the Fixed Effects (FE) Model is the best-performing model, particularly in larger datasets, as it minimizes both bias and RMSE. The Random Effects (RE) Model can also be effective, especially when collinearity is moderate and when the assumptions of random effects hold true. However, for datasets where collinearity is high, or where individual-specific effects are crucial, the Fixed Effects Model provides more reliable estimate. The Pooled Regression Model should generally be avoided in cases where collinearity or panel-specific heterogeneity is significant, as it produces the least stable and least reliable results across different collinearity levels