10 BULLETIN 897, U. S. DEPARTMENT OF AGRICULTURE. 
Since an average shipment contains only 1 package, it would be . 
difficult to select 50 in actual practice. Hence, 10 is used as arepre- 
sentative sample. Then the calculated maximum error on the. 
average weight of a representative sample is— 
4/(0.00)? + (1.00)? + (0.00)?-+ (.50)? + (1.00)?+ (.69)2+ 
+ (2.00)? + (4.00)? + (1.00)? + (4.00)? 
vid 
or 3.57 ounces. 
The same general scheme of calculation has been followed in 
Tables 5 and 6. The average number of packages per shipment was 
determined by inquiry at a large number of factories and inserted 
in round numbers. Any departure from the rules laid down are 
explained in the footnotes of the tables. 
EXPERIMENTS ON GOOD. COMMERCIAL PRACTICE METHODS. 
In order to try out the specified method of packing, a series of 
experiments was conducted in several factories. The method of 
packing was outlined to the packer, unless he was using it already, 
and care was exercised to see that all details were complied with and 
that a commercial speed was maintained. When everything was in 
good working order and the actual aim of the packer as to the weight 
he intended to put in the package had been determined, 50 packages 
were put up by him, after which they were carefully reweighed on an 
accurate scale to determine the actual variations. The results of 
these experiments are given in Table 7. 
Column 6 shows that some results are plus and some minus, that 
the averages of the various groups are near 0, and that more results 
are near their group averages than are far away from them. Prac- 
tically the same conditions govern normal frequency distributions; 
that is, if an infinite number of results are taken when the desired 
average minus aim is 0, the number of pluses and minuses will be 
approximately the same, the average will be 0, and comparatively 
few results will be far from 0. Then the groups are random selec- 
tions from normal frequency distributions, the precision measures of 
which are the figures under the designation, ‘‘ Average deviation from 
0 average minus aim,” given in the last column of the table. For pur- 
poses of comparison the practical limit of distribution of each group 
is considered to be the figure for the corresponding size given in the 
last column of Table 5 and of Table 6, headed ‘‘Calculated maximum 
error on the average of a representative sample.” 
It has been found that the weights of single packages taken from a 
single factory at one time follow a normal frequency distribution. 
The precision measure of each distribution of single packages from 
their averages is given by the figures in the last column of Table 7, 
‘Average deviation of single packages from average.’”’ For the sake 


— 
Date 
, 
