570 EXPERIMENT STATION RECORD. 



Among conclusions drawn are the following : 



" The kind of cattle should determine to a large extent the length of time to 

 feed." " Corn silage in all rations where it was fed proved a more economical 

 and a more profitable roughage than clover hay alone for fattening cattle." 

 " The addition of corn silage to a ration of shelled corn, cotton-seed meal, and 

 clover hay increased the rate, and decreased the cost of gain, and effected 

 equally as good finish on the cattle." " In order to induce sufficient grain con- 

 sumption to insure satisfactory gains it was necessary to limit the amount of 

 silage fed during the latter part of the feeding period." " The consumption of 

 frozen silage scoured the cattle." 



Sheep manag'ement, F. Kleinheinz (Madison, Wis., 1911, pp. XIII +225, 

 pi. 1, figs. 59). — A brief guide to practical problems in sheep management 

 which confront those engaged in sheep husbandry. The information is based 

 on the many years' experience as a shepherd of the author, who has had charge 

 of the flock at the Wisconsin Station since 1890. 



Mutton finishing on silage, J. E. Poole {Breeder's Oaz., 60 {1911), No. 

 25, pp. 1296, 1324, 1325, figs. Jf). — Details are given of the methods practiced 

 where feeding silage to sheep has been undertaken on a large scale. 



Studies on the strength and elasticity of the wool fiber. — I, The probable 

 error of the mean, J. A. Hill {Wyoming Sta. Rpt. 1911, Sup., pp. 139). — Con- 

 tinuing earlier work (B. S. R., 21, p. 73), the stretch of 5,000 and the breaking 

 strength of 59,400 wool fibers were measured. The fibers were taken from 26 

 fleeces, representing a wide I'ange as to character of fiber, breed of sheep, and 

 geographical distribution. 



In a single sample the breaking strain and the relation between breaking strain 

 and area of cross section was found to be more variable than the relation be- 

 tween the two characters, when subsamples of 100 fibers each were drawn 

 from unmixed samples. By this method, however, more than half of the means 

 of the hundreds differ from the means of the thousands by more than the 

 probable error of the means of the hundred. The variation of the means of 

 subsamples of 1,000 fibers were determined as drawn from unmixed samples, 

 but all of this work pointed to the conclusion that the probable error calcu- 

 lated by the formula 0.6745a-/V->i, where er is the standard deviation and n the 

 number of variants, is of little or no value as a measure of the accuracy of the 

 means obtained by the methods which hitherto have been used in this investi- 

 gation. It is thought that the fault may be that the groups of hundreds and 

 thousands studied have not fulfilled the conditions necessary to make them 

 random samples. The investigations were continued by changing the method 

 of drawing the fibers, but the study of deviations of the means in all cases 

 furnished evidence that the changed methods of drawing subsamples still failed 

 to follow the laws of the means of random subsamples. 



A final test was made to determine whether or not very small samples of wool 

 which contained barely the total number of fibers tested can be so mixed that 

 the accuracy of the means of thousands is described by the probable error cal- 

 culated by the standard deviation of the individual measurements, and also to 

 determine whether these means of thousands are accurate enough to be used in 

 measuring the effect of change of condition on the strength of wool. The re- 

 sult, however, was somewhat contradictory, and if it can be said to prove any- 

 thing at all, it is that mixing very small samples before testing can not be de- 

 pended upon to give means of thousands whose theoretical probable errors are 

 good measures of their accuracy. 



Tables are given which show the comparison of the theoretical with the true 

 probable errors of the means of hundreds for all the samples studied, and com- 

 parisons of the means of thousands for samples from which 5,000 or more fibers 



