366 EXPERIMENT STATION EECOED. [Vol. 41 



resting metabolism of a young woman exposed to inclement cold winds. A 

 calculation of the cooling power on the surface of the chock as determined by 

 the Kata-thermometer gave results from six to seven times that of the loss 

 of body heat, showing that under the conditions of cool moving air the resting 

 metabolism is apparently controlled by the surface temperature of the cheek 

 and can be calculated by taking the record of the dry Kata-thermometer cooling 

 power and cheek surface temperature. 



ANIMAL PROBTJCTION. 



The American Society of Animal Production. — Record of proceedings of 

 annual meetings, December, 1915, and December, 1916 (Amer. Soc. Anim. 

 Prod. I'rocs. lUl.5-16, pp. 210, figs. 30). — This is a combined report published 

 December, 1917, of the meetings held, respectively, at Manhattan, Kans., and 

 Urbana, 111. Some 27 papers reporting or discussing research are printed, to- 

 gether with a report on Methods of Experimentation in Animal Nutrition (pp. 

 101-111) by a committee composed of H. S. Grindley, "W. A. Cochel, J. M. Ev- 

 vard. L. D. Hall, and W. A. Carroll. 



The average correlation within subgroups of a population, S. Wright {Join: 

 Wash. Acad. Sci., 7 {1911), No. 11, pp. 532-537).— The "populations" used in 

 breeding experiments and similar investigations are frequently composed of 

 several fairly well differentiated groups, such as families or even breeds. A 

 measure of correlation of two characters in the entire population, disregarding 

 as far as possible the group differences, is often desired but, as the author points 

 out, the ordinary coefficient of correlation does not meet the requirements be- 

 cause its magnitude is influenced by heterogeneity in the material, which, more- 

 over, can seldom be considered " normal." The average of the coetRcients of cor- 

 relations of the individual groups is suggested as a good approximation to the 

 desired measure. If the separate group correlations are not needed for the 

 purpose of the experiment or their determination would be statistically worthless 

 on account of the small size of the group, the straightforward computation of 

 this average involves some very laborious numerical work not otherv/ise utilized. 

 Formulas developed in this paper indicate how the average group correlations 

 can be computed from a knowledge of the correlation and standard deviations 

 of the two characters in (1) the total population, and (2) the means of the 

 groups. In many cases all these statistical constants will have a value in con- 

 nection with an investigation apart from this use. 



Further illustrations of the applicability of a coefficient measuring the 

 correlation between a variable and the deviation of a dependent vai'iable 

 from its probable value, J. A. Harris {Genetics, S {1918), No. 4 pp. 328-352, 

 figs. 6). — In a previous paper ^ the author pointed out that in studying the inter- 

 relations of two variables in which one is a fraction of the other (such as the 

 number of fertile eggs laid by a hen and the percentage hatched) the ordinary 

 coefficient of correlation does not provide as much useful information as a co- 

 efficient showing whether the value of the fraction becomes relatively larger or 

 smaller A'iith increasing values of the independent variable. The formulas de- 

 rived in the earlier paper are repeated in a form convenient for computation and 

 supplementary ones added A number of illustrations of the use of the coef- 

 ficient, both with plant and animal material, are given, including problems of 

 fertility and fecundity, the proportionality of parts, the relationship between 

 chemical data such as total solids and a particular solid constituent, and the 

 measurement of differential viability of zygotic tj^es in a breeding experiment. 



iBlometrika, 6 (1909), No. 4, pp. 438-443. 



