]\gt ; Mathematical Contributions to the Theortf of Evolution .
48 was always less than the standard error when only was taken into account , unless We may now prove the similar theorem that when we use three variables , , on which to base the estimate , the standard erlOt will be again decreased , unless The condition that , in our present case , shall be less than in the last , is , in fact , This may be finally reduced to\mdash ; that is The treatment of the general case of val.iables , so far as obtaining the , is obvious , and ib is ) necessal'y to give it at length .
We can now see that the use of normal ssion formulae is quite in all cases , so long as the linlitatio]ls of interpretation are recognised .
remaillS a , of correlation .
These results 1 must plead as for my use of normal formulae in here the was markedly non-normal .
" " Mathematical Contributions io the of tion.\mdash ; On a Form of Spurious Correlation which may arise when Indices are used in the cnleIlt of \ldquo ; By KARL , University ceived ) , 1896 , \mdash ; Read , 1897 .
( 1 ) If the of two absolnte rements on the same 01 different organs be taken it is co1lvenient to terlll an index .
If and be two functions of the three variables , and these variables be selected at random so exists no elation between , or will still be ound to * Economic Journal , ' Dec. , 1895 .
and Dec. , 1896 , " " On the relation of Total Pauperism with Proportion of Out-relief.\ldquo ; Prof. Karl Pearson .
exist correlation between and .
Thus a real danger arises when a statistical biologist attributes the correlation between two functions like and to organic relationship .
The particular case that ir likely to occur is when and are indices with the same denominator .
for the colrelation of indices seems at first sight a very plausible measure of organic correlation .
The difficulty and danger which arise from the use of indices was brought home to me recently in an endeavour to deal with a considerable series of personal equation data .
In this case it was convenient to divide the errors made by three observers in estimating a variable quantity by the actual value of the quantity .
As a result there appeared a high .
of correlation between three series of abso .
lutely independent judgments .
It was some time before I realised that this correlation had nothing to do with the manner of judgin but was a special case of the above principle due to the use of indices .
A further illustration is of the following kind .
Select three numbers withiu certain ran at random , say , these will be pair and pair ] ated .
Form the proper fractions and for each triplet , and correlation will be found between these indices .
The application of this idea to biology seems of considerable importance .
For example , a quantity of are taken from an ossuarium , and ars put together in groups , which are asserted to be those of individual skeletons .
To test this a biologist takes the triplet femur , tibia , humerus , and seeks the correlation between the indices humerus and humerus .
He might reasonably conclude that this correlation marked organic relationship , and believe that the bones had really been put together ntially in their individual grouping .
As a matter of , since the coefficients of variation for femnr , ibia , and humerus are approximately equal , there would be , as we shall so later , a correlation of about to ; between these indices had the bones been sorted absolutely at random .
I term this a spurious organic correlation , or simply a spurious correlation .
I understand by this phrase the amount of correlation which would still exist between the indices , were the absolute lengths on which they depend distributed at random .
It has hitherto been usual to measure the organic correlation of the organs of shrimps , prawns , crabS , , by the correlation of indices in which the denominator represents the total body length or total carapace length .
Now suppose a table formed of the absolnte lengths and the indices of , say , some thonsand individuals .
Let an " " imp\ldquo ; ( allied to the Maxwellian demon ) redistribute the indices at random , they wouJd then exhibit no correlation ; if the esponding absolute lengths followed along with the indices in the redistribution , they also would exbibit no correlation .
Now let us snpposo the indioes not to have been ] ated , but the imp to redistribute the abso .
Mathematical ibutions to the Theory of Evolution .
Iute lengbs ; these would now exhibit no organic correlation , but the indices calculated from this random disbribntion would have a correlation nearly as high , if not in some cases higher than before .
The biologist would be not unlikely to argue that the index tion of the imp-assorted , but probably , from the vital standpoint , impossible beings was " " organic As a last illustration , suppose 1000 skeletons obtained by ibuting component bones at random .
Between none of their bones wilI these individuals exhibit correlation .
Wire the spurious skeletons together and photograph them all , so that their stature in the photographs is the same ; the series of photographs , if measure , will show correlation between their parts .
It seems to me that the biologist who reduces the parts of an animal to fractions of some one measured upon it is dealing with a sel'ies very much like these photograpbs .
A part of the correlation he discovers between is undoubtedly organic , but another parb is solely due the nature of his arithmetic , and as a measure of organic relationship is Returning to problem of the randomly distributed bones , let us suppose the indices femur/ humerus and tibia , humerus to have a of .
Now suppose CCeSSiVely 1 , 2 , 3 , 4 , \amp ; c. , per oent .
of the bones are assorted in .
true oupings , then the true organic col'relating of the .
It starts from and will alter gradually until 100 per cent. of the bones are grouped .
The final value may bo greatcr 0 less than , but it would seem that is a correct point to measure the organic correlation from than .
At any rate it appears fairly certain that if a biologist recognised that a perfectly random selection of organs would still lead to a correIation of organ-indices , he would be unlikely to accept index-col.relation as a fair measure of the rela- tive intensity of correlation between organs .
I shall ordingly define spurious organic corrcIation as the correlation which will be found between indices , when the absolute values of the organs have been selected purely at random .
In estimating relative correlation by the hitherto usual measurement of it seems to me that a statement of the amount of spurious col.relation ought always to be made .
( 2 ) Proposibion I.\mdash ; To find the mean of an index in terms of the , coefficients of riation , and coefficient of correlation of the two absolute measurements .
* Let be the absolute sizes of any four correlated organs ; their mean values ; standard deviations ; * In all thnt follows , unless otherwise stated , the correlation of any kind whatever , i.e. , the frequencies are not supposed to follow thw Oaussian or normal law of error .
Prof. Karl Pearson .
their coefficients of variation , i.e. , rsspectively ; , the six coefficients of correation ; the deviations of the foul .
organs from their means , ; the mean of the index , and the mean value of ; the standard deviations of the indices and respectively ; and the totaI number of groups of organs .
We shall suppose the ratios of the deviations to the mean absolute values of the are so small that their cubes may be neglected .
Then if we neglect quantities of the third order in .
But , and Hence : . . . . . . . .
( i ) .
Similarly .
( ii ) .
Thus we see that the mean of an index is not the ratio of the means of the corl.esponding absolnte measurements , but differs by a quantity on the correlation and variation coefficients of the absolute measurements .
( 3 ) Proposition II.\mdash ; To find the standard of an index in of the coefficients of , and coefficient of correlatiomJf the two absolute measurements .
square terms ) , if we neglect cubic terms . . . . . . . . . . .
: ( iii ) .
Nathematical tributions to the volution .
493 ( 4 ) Proposit the coefficient of of two in ients of the coctficients of correlation of the four absolute their coo .ffixients of variation .
Let and be the two indices .
if we neglect terms of the cubic order . . .
Hence , finally , . . . . . . .
( 5 ) Thus we have exprcssed in terms of the four coefficicnts of ation and the four coefficients of of the absolute measurements which form ) indices .
We mity draw the following clusions : The ] ation between two indices will always when the four absolute measurements the indices are quite uncorrelated ( ii .
) If two of the organs are ctly c , let us ical : for example , the and fourth , so that , and find ( v ) .
is the coefficient of correJation between two indic with same denominatol and .
The value of in ( v ) does not vanish if the be quite correlated , i.e. , .
In this case , . . . . . . . . . . . . .
( vi ) .
This is the measure of the spurious correlation .
For the Prof. oase in which the coefficients of variation are all the same , When the absolute sizes of organs are very feebly correlated , then in most cases there will be a considerable correlation of indices .
Example .
Suppose three organs , , and to have sensibly equal coefficients of variation , and that the correlation of and and of and , as well as of and Then : This formula illustrates well in a specially simple case how the in the indices diverges from the spurious value , as we alter and from zero , , as we introduce organic correlation .
According as , the correlation of the numerators , is greater or less than , the correlation of the numerator with the denominator , the actual index correlation can be greater or less than the spurious value . .
If bs the indices , then iu the case of normal correlation the contour lines of the correlation surface for the indices are given by constant , where , and are given by ( iii ) and above .
The contour lines of a surface of spurious index correlation are given by constant , while the uncorrelated distribution of the numerators and is . .
given by the We are thus able to mark the growth of the spurious correlation nstant .
as we increase from zero ; we see the axes of the ellipses diminishing and their directions inning to rotate .
Example .
To find the correlation between the two chief cephalic indices .
made on 100 ' ' Altbayerisch\ldquo ; 8 skulls , by Professor J. Rank .
See I have calculated the following results from the ements his 'Anthropologie der Bayern , ' Bd .
, Kapitel , S. 194 , Mathematical Contributions to the of volution .
495 Breadth of sknll:* Height of skull : ) Lsngth of skull : ephalic index , : Cephalic index , : ) , Cephalic index , : ) The coefficients of relation may at once be deduced : Breadth and length : Height and length : Height and breadth : This is the first table , so far as I am aware , that has been published of the variation and elation of the three chief cephalic lertgths .
It shows us that there is not at all a close correlation between these chief dimensions of the skull , and that a small factor for size is to be sought in the correlation of height and length , i.e. , while a broad skull is probably a long skull and also a high skull , sknll will probably be a short skull , and a low skull a long skull .
Without ting the values of in , we can find , or the correlation between length and indices from : .
This follows at once from the general theorem iven in my memoir on " " ression , Panmixia , and Heredity 'Phil .
Trans vol. 187 , , p. 279 , or by substitution of the above values of in ( v ) , we find : If we calculate from ( vi ) the correlation between the same cephalic in lices on the hypothesis that their heights , breadths and lengths are distributed at random , i.e. , that our " " imp\ldquo ; has constrncted a number of arbitrary and spurious skulls Banke 's measurements , we find : It seems to me that a quite erroneous impression would be formed of the organic correlation of the human skull , did we judge it by the magnitude of the correlation coefficient the two chief the bsolute measures givcn are in millimetres , and the coelficicnts of variation are percentage variations , i.e. , they must be divided by 100 before being used in formulae ( i ) , , and ( iii ) .
I hope later to correlation in man refelence to race , sex , and organ , as I have treated vnliation .
496 Prof. Karl Pearson .
cephalic indices , for no less than of this would remain , if we destroyed all organic relationship between the lengths on which these indices are based .
' Example ( d ) .
To find the spurious correlah on between the indices humerus and made boganei oetons.ittheilungen aerThe sowing rfrom measurements medicinischen Facultat der K. J. Universitat , Tokio , ' .I .
Tables .
) I have kept the sexes apart although there are but few of each .
8 Skeletons .
Number to 44 .
Measurements in centimetres .
Femur , : Tibia , : Humerus , : The following coefficients of colrelation were calculated directly : Femur and tibia : Femur and humerus : Tibia and humerus : From these were deduced by the formulae of this Index , : Index , : Index , : Hence we find for the correlation of the indices and : But the spurious correJation , if the bones had been grouped at andom would have been have to thank Miss Alice Vee for a considerable part of the arithmetic work of this example .
values for the indices are not in absolute agreement with those to be deduced from the lengths , for it was not always possible to use same skeleton for femur and humerus as for tibia and humerus , i.e. , sometimes one or other bone was missing .
For same , the constants for the absolute lengths do not entirely with those given for Ainos in my paper on " " Variation in Man and p. 303 ) , for the simple renson that I there used every available bone , and not every Woman\ldquo ; in The Chances of Death and other Studies in Evolution , ' vol. 1 , available pair , as here .
to the bulating torres olltibies f the obhel we find ( Skeletons .
Numbca ) to 24 .
AIcasurements in enlur , : Tibia , : ) ) .
Humerus , : ) , and tibia : and humerus : Tibia and humerus : : : : Hence we may conclude as follows : ( i ) The absolute lengths of the bones differ from hose the in being very closely COl.related .
( ii ) The use of indices for the long bones would appcal ' to mlse , rather than , as in the case of the skull , to COl.relation .
( iii ) measure , howevelnnic corrclation of tho indices ) , we find index ] ation less ] for both lon bones and skull , and in both the comparatively small as compared with bhe latter .
( iv ) The results for the 24 skeletons , based ( 11 but few dar.a , serve on the whole to conIirm the male resuIts .
* ( 6 .
) From the above examples it will be seen that the method , judges of the intensity of relation by the l.eduction all absolute measures to indices , the denominafol.s of which ] some bsolute mement , is not force .
for this method would give ion of the lation had the of the been togethel at random , , if there were no correlation at all .
TJIe follo ing additional remarks may ] ) of .
The show us that the elation coefficients of indices are not only of the correlation coefficients of of the coefficients of variation of the latter surements .
Hencc , * The { is more in itting , in fetnul , } in than the female , she b than is iuRtatul.e , to in Womun ppcars tound tated itature ) Galton .
to the Memoir the coefficieIlts of variation be constant for local races , it is impossible that the coefficients of correlation catl be constant for indiccs .
In other word , the hypothesis of the constancy for local races of correlation , and that of the constancy for local races of Yariation , stand on exactly the same footing .
The conclusions of this paper although applied to organic tion are equally valid so far as concerns the use of indices in judging the ation of either or economic phenomena , .
It was , indeed , a difficulty arising from my discussion of personal \mdash ; a spurious correlation between the judgments of different relS \mdash ; which first drew my attention to the matter .
Notc , , 1897.\mdash ; The result described by Pearson evidently affects the value of the correlation coefficients determined by me in rangon and reinus ( ' Roy .
Soc. Proc vols .
and 54 ) , becanse I have always expressed the size of the organs measured in terms of body length .
In order to the effect of this , I lately performed , at rofessor Pearson 's gestion , the followirg experiment : It happens $.hat my measures of vmouth shrimps are recorded in a book , in the order which they were measul.ed , and therefore at random as egards carapace length or other characters .
I constructed from less records 420 " " spurious shrimps , in the following way : length of ) first shrimp in the book was associated with the ClI'apace length of the tenth shrimp and the " " post-spinous lengbh\ldquo ; of the twentieth , and so throughont .
Evident ] whese three measures were associated at random , and we might expect tltat these spurious shrimps would show no organic correlation ; but when the carapace and " " lengths\ldquo ; of tlJese spurious shrimps were divided by the body , and correlation between the resulting indices was determined , the value of was found to be the value for real shrimps , or the correlation due to the use of indices forms 47 per cent. of the observed value .
W. F. R. WELDON .
Notc to the Memoir Profcssor Karl Pearson , F.B.S. , , purious \ldquo ; By FRANCIS GALTON , F.B.S. Received ebruary 1 , 189 I sond this note to serve as a kind of ) ppendix to the memoir of Professor K. Pearson , elioving that it may be useul in enabling to realise the { lesis of spurious corr e. It is ) rather difficult to do so , because the results rived at iu the , which of serious interest to practical statisticians , ve firsb a

