]\gt ; Mathematicat Contributions to Theory of Evotution .
Calculated as above the composition of the undilubed black-damr } was thus : . . . . . . . . . . . . . .
Carbonic acid . . . . . . . . . . . .
14 70 This result differs little from that given by the other samples .
II .
" " Mathematical Contributions to the Theor .
of Evolution .
II .
Skew in Homogeneons \ldquo ; By KAItL PEARSON , University College , London .
Communicated by Professor RICI , .
Received December 19 , 1894 .
( Abstract .
) In the deduction of the nol.mal cve of equency f the symmetrical point binomial , three conditions are usually assumed to be true : The chances of any " " contributory cause " " giving its unit of deviation in exc.ess or defect are presumed to be equal .
The of " " contl'ibutory causes " " are supposed to be indefinitely great .
The " " contributory causes\ldquo ; are all supposed to be independent .
( c ) amounts to the assumption of a binomial form , ( a ) to the equality of and indefinitely great value of It is in the papel ' that there is an important geometrical relation between the normal frequency curve and the symmetrical point binomial , , which is trne pendently of the magnitude of .
Thus the condition ( b ) is not necessary to the very close fitting of point omials to normal curves for eveu very small values of , such , for example , as 8 or 10 .
This has been long recognised statistical practicc if source has not been noted .
We can ) emove condition ( a ) from our a priori limitations by finding a curve bich is related to the skew binomial in preclsely the manner as tho norm curve is elated to the metrical binomlal .
The equation to this is This relation must be satisfied or nearly satisfied if a series of vations or measurements is to be fitted with the skew curve , whioh related to the skew point-binomial as the normal curve to the metrical point-binolnial .
fitting a skew point-binomial mUbt have .
For the normal curwe .
But a great number of \mdash ; especially in anthropometry and zoometry\mdash ; give .
Hence they differ frovn the llorma , I curve in the opposite direction } the skew poiut-binomial and its corresponding frequency curvs .
After the complete theory of the fitting of skew this special skew bas been discussed with exampies , the memoir proceeds to the isation of the frequency lrve by withdrawing limitation ( c ) above .
Just as the symmetrical nomial and normal curves are illustrated by the tossing of a group of coins , and the skew binomial and its skew curve by the ning of a group of -sided teetotums , so we can arrive at a riss of curves in which the contributory causes are interdependent by considering the withdrawal of cards from a pack of ns cards taimng suits ; or , again , by drawing a definite amount of sand a vessel containing two kinds of salld .
For discontinuous series the solution is a hypergeometrical seriu .
If now a curve be formed which is related by the same fundamonbl yeometrical relation to this hypergeometrical sel.ies as the curve to the symmetrical point-binomial , or the first skew the skew point-binomial , we obtain a generalised equency o which contains both those hitherto considered as special or limibl cases .
It is not suggesed that the hypergeometrical series or its sponding curve is the only case in which the a pr.iori condition ( l ) dependence of " " contributory causes " " is replaced by an pendence .
But it is suggested that it is of the most cases , and one which naturally occurs at the commencement investigations .
That it is probably quibs sufficient is evidenod..S homogeneous atatistics which cannot blosely exprdttle futhor hitherto failed t by the curves which correspond to the hypergeometrical series ibutions to the Theory of Evolution .
The differeutial equation to the generalised frequency curve is shown to be of the form If we put we have the curve corresponding to the skewbinomial ; if we put we have the normal curve .
In the most general case arc led to two principal types of curves The second of these curves is marked by a limited range and skewness .
Its theory\mdash ; method of fitting to actual statistics and its geometrical properties\mdash ; are discussed , and the curve is shown to involve in fitting only the use of a table of -functions\mdash ; a table which already exists .
The first of these curves has skewness but no limit to ) .
This unlimitedness of range is not , however , necessarily nificant .
There is a limit to the height of adult males , or at any rate to the ratio of their sitting to standing , but we do not hesitate to express esults in terms of the normal curve .
The fact is that both curve and generalised cllrve close approximations to series\mdash ; point-binomial and point-hypergeometrical series\mdash ; which can themselves give a limited range , and ought to fit these selies rather than the curves to our The criterion to distinguish between the application to any speclal case of curves ( i ) or ( ii ) negative or positive value of 2 .
which we have seen vanishes for the curve corresponding to the skew point-binomial .
The complete eatment of es of the first kind is shown to depend on a called a ) -function .
This G-function has been discussed in a recent paper by Dr. Forsyth , to whom the author had referred for information with regal.d to it .
It is built up of -functions with imaginary arguments .
The function has not yet been tabulated , ] ) various nlulae are given for evalnation , and it is hoped its values may shortly be calculated for the range * The fltting of the first series is discussed in is memoir ; the fitling of geometrical series is selved as ) memoir is alrcady of considerable species of skew curve , he has hitherto failed to find one which distingDish a compou1ld curve , eneous m skew resulting from skew varintion homogeneous matorial .
He does not despair , however , of ultimately finding suctl a rion .
The test of actual fitting is generally sufficient , but , of 1nborious .
PART II.\mdash ; Illustration .
The second of the memoir the minimum of tion , which the author considers absolutely necessary , in order tp demonstrate that the generalised curves reached are apable of widest application to every variety of practical statistics .
illustrations- show that from the slight amount of skewness usualIy neglected by statisticians\mdash ; although of vital import when coml to consider variation with growth , as in statistics of iation with growth\mdash ; even to extreme cases in which the curve is asymptotia to the ordinate of maximum frequency , a good frequency curve can be found .
Although the number of illnstratior is considerable , and is only a part of the author 's , yet hesitates at esent to make any dogmatic statements with rogard to the relations of skewness in variation to secular evolution ; but ho believes that the persistent ecurrence of certain types of curves in zoometry and of directions of skewness in anthropometrio statistics will be found , as sufficient accumuIates , to broad generalisations , although at present they can only be treated al buggestions for further investigation .
The special illustrations given are : barometer 1-ariation variation in crabs and prawns ( Weldon ) , in height of American cruits ( Baxter ) , American school girls ( Porter ) , in length-breadth index of Bavarian skulls ( Rank ) , frequency oenteric fever politan Asylums Board ) , guesses at tints ( Gresham College ) , divowo statistics ( tVillcox ) , variation in house-value ( Goschen ) , variation in buttercups and clover ( De Vries ) , ariation in pauper ( Booth ) , and resolution of the ghsh male mortality curve into skew components .
The memoir concludes with some general remarks on the modiflW tions required in the theory of correlation by ) use of curves , but reserves for the present its complete discussion .
British Association have kindly given a gnnt for this purpose .
It is noteltorthy that of compoundedoess dealt with hithorto by nthor givt positive .

