492 



Prof. Karl Pearson and Miss Alice Lee. 



tinuously from this type as we pass from station to station, and 

 appears to be fairly uniform along lines which are termed gener- 

 alised isobars. If this result should be confirmed, then the positions 

 of these generalised isobars would require a larger series of stations 

 with observations for a longer period of years to determine them 

 accurately. 



3. The authors' next object was to discover what constants of the 

 barometric frequency suffice to describe it with the least probable 

 error. A somewhat elaborate investigation was accordingly made 

 into the probable errors of the constants, and four physical quantities, 

 the mean, the variation (or standard deviation), the skewness, and 

 the modal frequency were found to be the constants, which described 

 a local barometric frequency with the smallest probable errors. 



4. They have next discussed the chief physical features of a 

 barometric frequency distribution. 



(a) The modal height and the modal frequency are found to 

 possess certain advantages over the mean height and the mean fre- 

 quency. Various methods are considered for approximately deter- 

 mining the position of the mode. The mean modal height for 

 Great Britain is shown to be very nearly the standard atmosphere 

 of 30". 



(6) The variability of the barometric height and the skewness of 

 the distribution are discussed at some length. A formula is given 

 in terms of the skewness for calculating the probability that the 

 barometer at any station will be in excess or defect of the mode or 

 of the mean. 



(c) A test of the accuracy of the observations for the twenty coast 

 stations is made by attempting to interpolate the frequency constants 

 of London and Cambridge from those of Southampton, Hillington, 

 and St. Leonards. The results are fairly close when tested with the 

 results for London and Cambridge calculated directly from the 

 telegraph station returns. 



5. In the second part of the memoir the subject of correlation is 

 dealt with. It is shown that within the limits of the British Isles 

 there is a very high degree of correlation (as high as 0*9824 between 

 Babbacombe and Churchstoke), only sinking to 0*7572 if we take 

 practically the utmost reach of the British Isles. The correlation 

 differs with directions more or less along and more or less perpen- 

 dicular to the generalised isobars, and it is pointed out that this 

 gradual change with direction and distance, combined with change 

 owing to the interval of time between observations, enables the 

 meteorologist to find systems of stations with almost every variety 

 of correlation coefficient. 



The probable height and probable deviation from that height at 

 any given station, based on a knowledge of the heights, contempora- 



