TABLES TO FACILITATE THE COMPUTATION OF THE 
PROBABLE ERROBS OF THE CHIEF CONSTANTS OF 
SKEW FREQUENCY DISTRIBUTIONS. 
By a. IIHIND, B.Sc, Biometric Laboratory, University College, London. 
The general theory of the probable errors of the ct)nstants of skew frequency 
distributions was originally given with illustrations by Pearson and Filon, Phil. 
Trans. Vol. 191, A (1898), pp. 229—311. The values there deduced depend on 
the form of special frequency curve adopted, and involve considerable arithmetical 
work for each individual case. In these frequency investigations the fundamental 
constants are the well-known /Sj and fi.,. Every frequency character expressible in 
terms of /3i and /So can have its probable error determined, provided we know the 
probable errors of ySj and /So and the correlation in deviations between and /So. 
General expressions for the probable errors (or the S.D.'s S/s,, of /Sj and 
as well as the coi-relation -R^,^, of deviations in ySj and /Sa, together with the 
probable error of the criterion (or its S.D. = S^^) were first given by Pearson, Pliil. 
Trans. Vol. 198, A (1902). These involve a knowledge of /3„ /S, and /S.*, 
further constants of the distribution which can only be found if the numerical 
values of /t,,, /ij and /Xg have been in some way determined. Now it has been 
shown that with the total frequencies usual in practice these high moments are 
subject to very large percentage errorsf, rendering their use extremely undesirable, 
even if we could overcome our natural repugnance to the great labour of calculating 
them. 
But we have to bear in mind that the exact value of a probable error is not 
usually desired. What we more often require is a rough determination of its 
t Pearson: "On the General Theory of Skew Correlation and Non-Linear Regression," Drapers' 
Company Research Memoirs (Dulau & Co., 1905), p. 8. 
