Data Type Detection for Choosing an Appropriate Correlation Coefficient in the Bivariate Case

The data scientists usually define a data type based on a nature of variables and select an appropriate correlation measure. However, this is not convenient and very time-consuming in data intensive domains. I propose to detect the types of variables and choose the appropriate correlation coefficient in order to automate the statistical procedure of correlation estimating from mixed data. This should lead to a reduction of time spent on correlation analysis and to increase the accuracy of estimation of correlation coefficients. The continuity index is used to detect whether a variable is continuous or ordered categorical. Based on simulation study I have estimated the cutoff level for the continuity index to choose the Pearson correlation, the polychoric, or the polyserial correlation coefficient.