ON DIFFERENT METHODS OF ESTIMATINGCORRELATION COEFFICIENTS

Abstract

Correlation analysis uses different methods to measure the extent of relationship between two variables. Different methods of estimating correlation coefficients have been used by various researchers, but the
challenge has been using the appropriate methods. In this study, fourteen different methods of estimating correlation coefficients were examined with emphasis on their assumptions, properties and relationships
existing among some of the methods. While derivations were made of Pearson’s correlation with its limit; Spearman’s correlation; Biserial and Point Biserial, Kendall’s Tau by neighbor swaps was illustrated with
order of performance of numeric data. The relationships among the methods examined showed that Phi coefficient has similar relation with Tetrachoric as Point Biserial relates with Biserial correlation coefficient. It was also established that Pearson’s correlation assesses linear relationship between variables, but Spearman’ rank correlation assesses monotonic relationship. Statistical applications of the methods to real life situations were also established.