A simple graphic method to assess correlation
- Biometrics & Biostatistics International Journal
Othmar W Winkler
PDF Full Text
The graph of two-dimensional data from the field of education, grades received and attendance at classes, of 115 observations, resembled a rectangular triangle, formed by the X and Y axes, and a line fitted to the highest data points as the hypotenuse. Taking advantage of its triangular nature, this study explores a simple way to assess the co-relation between grades and absence from classes yielding r²=84% a coefficient of correlation compared to the standard measure of association, yielding the very different correlation coefficient r² of only 17.5%.. The linear regression-correlation model was not appropriate for the data of this study. Instead, the proposed simpler method treats each data point as the factual reality, not as a deviation from a hypothesized ideal regression line. Given that the data were in the shape of a rectangular triangle, to assess the degree of association, between the two characteristics of the data allowed taking advantage of the geometry of rectangular triangles.
simple graphic method, regression/correlation analysis, descriptive statistics