88 MASS. EXPERIMENT STATION BULLETIN 251 



TECHNICAL APPENDIX 



In the foregoing- part of the bulletin, the emphasis was placed on the 

 presentation of the results secured from the study rather than upon the 

 statistical method used in arriving at the results. Son.e readers may be 

 interested in knowing more about the methods used in the analysis. For 

 their benefit the following teclinical section is included. 



For the statist'cai an-ilysis the factors i)revioiisly discussed were ex- 

 ]iressed in the following terms: 



A. Index of change in flock size, in terms of tiie ])ercentage which 

 the ;'.verage number of birds is of the initial number. 



B. Annual egg production i)er bird; the sum of tlie monthly 

 averages for each flock. 



C. Price of eggs; the sum of tlie monthly averaae prices for 

 each flock, divided by 12. 



D. Fall egg i)roduction in the four months Se])tember through 

 December, exjiressed .is a i^ercentage of tiie total jiroduction 

 for the year. 



E. Other credits than market eggs; credits from all sources ex- 

 cept market eggs, expressed in dollars per bird. 



F. Size of flock; the November 1 inventor^' figure for all females 

 at the beginning of the record year. 



G. Labor return from the entire flock. 

 H. Labor return per initial bird. 



Three i)roblems were worked tlirougli. Tiie first one took labor returr. 

 from the entire flock in 1927 as the dependent variable. Because of the 

 great differences in size of flock, as shown in Table 16, this factor was 

 so imjiortant as to overshadow most of the others. Consequently it was 

 thouglit best in the other analyses to eliminate the factor of size of flock 

 from the problem and to make labor return per bird the dependent factor. 

 This reduced the total correlation but brought out more clearly the re- 

 lationshiji between the other factors in^■<llved. 



Table 1(). Poultry- Flocks Classified According to Iiiit'al Size. 



Initial Number of Number of Flocks Studied 



Birds per Flock 1926 1927 



0— 99 23 20 



100 — 249 51 34 



250 — 499 33 36 



,500— 749 15 22 



750 — 999 11 8 



1000—1249 2 4 



1250 — 1499 2 6 



1500 — 1749 1 1 



1750 — 1999 3 2 



2000 and over 3 2 



Total 144 134 



The probable errors of estimate in Table 17 are valuable in showing 

 the closeness with which the labor returns could be estimated in the re- 

 spective years. The measure tells something about hov.' much was left 

 unexplained. The probable error of estimate for labor return per b=rd 

 in 1927, for example, means that by use of the regressions we were able 

 t(! estimate the l;d)or returns and have half of the estimates fall within 

 7S cents of the actual labor returns. 



