640 PROCEEDINGS OF SECTION G (l.). 



deception, resulting in disappointment to those erroneously informed, 

 and may finally recoil disastrously on the cause it was intended to 

 benefit. In any case it is detrimental to the cause of truth. 



The third consideration of importance is the necessity of caution 

 in drawing conclusions. Nothing is easier than to make mistaken 

 inferences from insufficient data, a practice frequently indulged in by 

 persons whose position in life makes it desirable for them to pose as 

 authorities. The earnest student of statistics, knowing the many 

 possibilities often involved, wiU as a rule proceed with extreme wariness 

 and not pronounce an opinion unless the case is absolutely unequivocal. 

 He will often confine himself to pointing out what appear to be all 

 the possible solutions of a problem without expressing any predilec- 

 tion for any of them. 



That, however, even statisticians do not always adopt the very 

 best methods available to attain practical results might be proved 

 readily enough. Take the compilation of vital statistics, where it is 

 a practically universal custom to collect and tabulate the figures, not 

 only for years but also for separate months, so as to obtain an indica- 

 tion of the fluctuations of births, deaths, and marriages in different 

 seasons of the year. The object is, of course, an excellent one, and 

 might be eminently useful to science. Unfortunately the method 

 adopted for collecting the information is everywhere, as far as I have 

 been able to ascertain, that of tabulating the information according 

 to the date of registration, and not that on which the events really 

 took place. As registration is not, as a rule, effected till some time 

 after the event, and in many cases this time may amount to weeks 

 or even months, it is obvious that the incidence of births, deaths, and 

 marriages is almost invariably removed from the month to which it 

 rightly belongs, and that any conclusions drawn from the fluctuation 

 of the figures are subject to this serious objection. The reasons for 

 preferring the date of registration to that of the actual event are, of 

 course, well known ; but whether, in view of the consequent invalida- 

 tion of the results for strictly scientific investigation, they are sufficiently 

 weighty, is a matter on which considerable doubt might be entertained. 



Closely connected with the third phase of the use of statistics is 

 the fourth, that of prognostication. Here it is obvious that the dangers 

 of error exist in still greater abundance. For not only is it necessary 

 now to be sure of all the conditions that determined the past, but the 

 possibility of their modification or even complete change has to be 

 kept carefully in view. 



I cannot resist the temptation of rescuing from the unwritten 

 annals of statistical history the following amusing incident. Several 

 years ago an over-sanguine statesman in a rising country instructed 

 the Government Actuary to write a forecast of the increase of popula- 

 tion for the next 10 years. An interesting little pamphlet was com- 

 piled, with a carefully drawn curve representing the probable increase 

 of the population during the ensuing decade based on recent past 

 experience. The pamphlet was duly boomed and forgotten ; but about 

 five years afterwards the chief of the statistical office in that country 



