CORRMAT/PROB has been written to create a correlation coefficient matrix (Q-mode or R-mode) from data containing random gaps. The coefficients then are tested against critical values at the 95 and 99% significance levels for each intra-pair comparison, the number of samples/variables in such a comparison being influenced by the number of data gaps. A linear percentage variation matrix also is created. Description and application of CORRMAT/PROB is discussed using an example to illustrate its flexibility and robustness in data with a range of varying gaps. © 1982.