Canonical Correlation Analysis - An Application to Weather Conditions Data

Gyasi-Agyei AK, Mintah AE, Abraham AY and Mustapha AM

Published on: 2025-05-01

Abstract

This paper aims to evaluate the degree of correlation between Ghana's heating and cooling factors, spanning the period from January 2000 to December 2024. By identifying the linear combinations of the two sets of data with the highest correlation, the study attempts to determine the correlations that currently exist between the two sets of variables. It uses a type of multivariate analysis called canonical correlation analysis, a widely used covariance analysis method that can be effectively partitioned into two subsets of response and predictor variables. The first canonical variate had canonical correlation coefficient of 0.658 which is statistically significant with F (9, 716) = 25.246 at alpha level of 0.05, followed by the second canonical correlation coefficient of 0.277 with F (4, 590) = 7.926, and the third canonical correlation coefficient of 0.157 with F (1, 296) = 7.473. The results rejected the null hypothesis and showed that the three canonical correlation coefficients are not equal to zero. The first canonical root is used to show that there is highly positive correlation between the heating variables and the cooling variables of weather conditions. Policy recommendations include adopting climate-resilient agriculture, expanding renewable energy, integrating green infrastructure in urban planning, enhancing meteorological systems, and promoting public awareness to support climate adaptation and sustainable development in Ghana.