504 



SCIENCE 



[N. S. Vol. XLII. No. 1084 



Card Used ^ = 3i iV = 23 



New York City and Official 



score 770 ± .047 .713 ± .069 



New York City and Cor- 

 nell score .719 ± .056 .803 ± .050 



Official score and Cornell 

 score ,831 ± .036 .769 ± .057 



These results for total score by the three 

 different cards are high, but they fall far 

 short of the r^l which indicates perfect cor- 

 relation, and which should be obtained if (a) 

 the score cards were perfect descriptions of 

 the barns as places for milk production, and 

 if (h) the inspectors had perfect judgment in 

 the filling out of the cards. It is interesting 

 to note that the grades assigned by the three 

 methods agree equally closely (within the lim- 

 its of the probable errors of sampling) what- 

 ever cards are used: New York City and 

 Official, New Tork City and CorneU, and Offi- 

 cial and Cornell gradings have sensibly the 

 same correlation. 



In the case of the New Tork City and the 

 Official score, methods are graded separately. 

 The correlations are 



For 34 barns, r^.480it.089 

 For 23 barns, r = . 412 ±.117 



Thus the correlation for methods are but 

 slightly over half the size of those for total 

 score. This suggests that the 40 points 

 awarded for equipment in both the New Tork 

 City and the United States Department of 

 Agriculture (" Official ") cards is the great 

 factor in bringing about a close agreement in 

 the results (total score) obtained by the two 

 methods. Correlating points assigned for 

 equipment only (total score minus score for 

 methods) I find 



For 34 barns, r = . 733 ±.054 

 For 23 barns, r=. 685 ±.075 



The lower correlation of the values assigned 

 for methods as compared with those for equip- 

 ment is perhaps the most serious criticism to 

 be made of the score cards. 



While the agreement between the scores as- 

 signed by different cards falls far short of per- 

 fect correlation, it is interesting to note that 

 the agreement is actually closer than that for 

 iacterial count in evening and morning milJc 



from the same dairies. Here the correlation 

 is only 



r = . 456 ±.111. 

 The relationships between the scores as- 

 signed by the various cards and the bacterial 

 counts are naturally the results of the great- 

 est interest. All the possible correlations have 

 been worked out and are presented in the ac- 

 companying table. 



COEEELATION OP BACTERIAL COUNT WITH SCORE-CARD 



ENTRIES 



Evening Milk, Morning Mllfc, 



Card Used 34 Barns 23 Barns 



New York City 



Total score ... — .077 ± .115 — .046 ± .140 



Methods only.. -|- .024 ± .116 — .061 ± .140 



Equipment only — .170 ± .113 — .022 ± .141 



Official 



Total score ... — .003 ± .116 — .196 ± .135 



Methods only. — .140 ± .113 — .310 ± .127 



Equipment only -|- .065 ± .116 — .039 ± .141 



Cornell — .013 ± .116 — .102 ± .139 



Twelve out of 14 of these correlations are 

 negative in sign. All six of those for total 

 score are negative. In other words, as the 

 ratings assigned by the inspectors became 

 higher the bacteria became fewer. This is of 

 course as one would like it to be. Practically, 

 however, the author's statement, that there is 

 apparently no correlation between the bacterial 

 count and score as expressed by any one of the 

 cards, fully expresses the facts. The constants 

 are almost without exception very small in- 

 deed. There is not a single one which can be 

 safely considered as significant in comparison 

 with its probable error ! 



Such are the results : 



a. The correlation between the total scores 

 assigned the same barns by the same inspector 

 using the three most important cards is only 

 about three quarters of its theoretical maxi- 

 mum value. The correlation between the 

 scores for methods only is less than half its 

 theoretical value. 



h. There is practically no correlation at aU 

 between the scores assigned the barns by dairy 

 inspectors and the bacterial content of the milk 

 which they place upon the market. 



c. When correlations so low as those deduced 

 from the present figures are found between 

 the bacterial counts of morning and evening 



