Composite Samples of Milk. 173 



the total amount of fat by the total quantity of milk, 

 making 4.22 per cent., instead of 4,20 per cent., which is 

 the arithmetical mean of the five tests. The quantities 

 of milk in the various lots do not enter into the calcula- 

 tion of the latter.* 



198. The second example represents more nearly 

 than the first one the actual conditions met with at 

 creameries and cheese factories. As a rule, the mixed 

 milk from a herd of cows does not vary more in total 

 weight or tests, within a short period of time like one 

 to two weeks, than the fibres given in this example. 

 On account of this fact, samples taken, for instance, 

 with a small dipper may give satisfactory results to all 

 parties concerned. If the different lots of milk varied 

 in weight and test from day to day, as shown in the 

 first case, it would be necessary to use a "milk thief" or 

 one of the sampling tubes for taking the composite 

 samples; the size of each of the samples taken would 

 then represent an exact aliquot portion of the various 

 lots of milk (182). 



199. A patron's dilemma. The following incident will fur- 

 ther explain the difficulties met with in calculating average tests 

 nf different lots of milk. 



The weekly composite sample of the milk supplied by a cream- 

 ery patron from his herd of 21 cows tested 4.0 per cent. fat. 

 One day the farmer brought to the creamery a sample of the 

 morning's milk from each of his cows, and had them tested; 

 after adding the tests together and dividing the sum by 21, he 

 obtained an average figure of 5.1 per cent, of fat. From this 

 he concluded that the average test of the milk from his cows 



' In tjie eipcrlment given on p. 14S. the nrithmptical mean of the 

 tests Is .'i.lS per cent., while the true average (at content of milt 

 Is 4.85 per cent. 



