Abstract:
Nowadays, one of the common connection types in statistical data processing is the connection between numerical (nominal) and binary variables. We propose a new coefficient for evaluation of this type connection. It is quite different from Pearson's correlation coefficient and focused at the “ledge” type of connection, which means that the binary variable is expected to take one of its values only within some unknown but fixed boundaries for the the numerical variable. We call it ledge-coefficient. Application scope for the proposed coefficient is discussed. Algorithms for calculating ledge-coefficient and for searching all cases when the ledge connection is extremely weak or strong are proposed. Comparative examples for estimating connection power by ledge-coefficient and by Pearson's correlation coefficient or by Kendall's tau are shown.
Keywords:binary and nominal variables, normative intervals, ledge connection, binary string.