A method is presented for fitting chi-square (or gamma) distributions to observed distributions, either positively or negatively skewed, of telecommunication variables such as critical frequencies and maximum usable frequencies. These observed distributions have been recorded for years only in terms of the 10, 50, and 90% points, but in applications it is necessary to estimate the entire distribution. It is assumed in the present study that the observed variable can be approximated by a linear function of a chi-square variable with unknown degrees of freedom. The three unknown constants are determined from the three per cent points and a convenient table. Several examples of complete observed distributions are compared graphically with normal, chi-square, Gram-Charlier, and Edgeworth fitted distributions.