Assessing Multiple Regression Analysis (MRA) model fit for forecasting air traffic movements using log transformation: a case study on ATNS air traffic movement dataset during COVID-19 pandemic
Loading...
Date
2024
Authors
Masekoameng, John Lehlaka
Journal Title
Journal ISSN
Volume Title
Publisher
University of the Free State
Abstract
The COVID-19 pandemic introduced unprecedented challenges to the aviation industry, significantly impacting air traffic movements (ATM). This study investigates the effectiveness of log transformation in evaluating the goodness of fit of multiple regression models in predicting ATM within the South African aviation sector. Specifically, it compares the performance of a standard Multiple Regression Analysis (MRA) model with a log-transformed MRA model to determine whether log transformation enhances model accuracy and reliability.
The research explores traditional model fit assessment techniques, including R-squared (R²), Adjusted R-squared (R²adj), p-values, F-tests, residual analysis, Mean Squared Error (MSE), and normality tests such as the Shapiro-Wilk Test. Using data from Air Traffic and Navigation Services (ATNS), the study applies MRA to assess the impact of key predictors such as revenue, lockdown levels, confirmed COVID-19 cases, COVID-19-related deaths, exchange rates, GDP, and population on ATM.
Findings indicate that the standard MRA model outperforms the log-transformed model in terms of explained variance, predictive accuracy, and coefficient significance. While the log-transformed model offers slight improvements in residual normality and insights into non-linear relationships, it does not surpass the standard model in overall predictive power. As a result, the study concludes that, for practical forecasting and decision-making in air traffic management, the standard MRA model is preferable. However, future research exploring non-linear relationships may benefit more from advanced modeling techniques, such as polynomial regression or machine learning, rather than a simple log transformation.
Description
Dissertation (M.Sc.(Applied Statistics))--University of the Free State, 2024
Keywords
Multiple Regression Analysis (MRA), Log transformation, Air Traffic Movements (ATM), COVID-19, Aviation forecasting, Model fit assessment, R-squared, Adjusted R-squared, Residual analysis, Statistical modeling, Air Traffic Navigation Services (ATNS), Normality tests