An Approximate Statistical Test for Correlations between Proportions

The observed means and variances of data occurring as proportions or percentages may be used to estimate analogous parameters of a theoretical open array, X, which, on closure, yields a new array, Y, whose means and variances are exactly those of the observed data, but in which the covariances have been generated entirely by closure. The correlations in Y, found directly from the means and variances of X, are appropriate null values against which the observed correlations may be tested, A testing procedure is outlined, and a practical example is given.