Matrixboot: A Fortran Program to Bootstrap Pearson Correlations Computed from Data Matrix Input

Researchers have demonstrated that the bootstrap method of nonparametric inference is an acceptable alternative for conducting significance tests for correlations when the usual parametric procedures are not warranted. MATRIXBOOT is a FORTRAN program which computes and bootstraps correlations using two square-data matrices as input. The program constructs a distribution of 1000 correlations and computes summary statistics on the distribution. Summary statistics include the mean, standard deviation, and median. The program also computes and reports both 95% and 99% confidence intervals using an algorithm developed by Strube in 1988. The method is less subject to Type I error than are previous methods.