10 CIRCULAR 816, U. S. DEPARTMENT OF AGRICULTURE 
fixed representation to parts near the center and the edge. A less exact 
plan of taking the samples over a large part of the field for example, an 
arc cutting well into the field and returning to the margin at another 
point—will do as well. 
Sampling in the field margin may be carried out as the worker returns 
to the car. It would seem feasible to cut down marginal sampling. Table 6 
shows that marginal populations had but little effect on weighted aver¬ 
ages. While they are higher and more variable than field populations, the 
margins constitute only a small proportion of the total area in the northern 
Great Plains. Taking only two 1-square-foot units in the margin will not 
mean much sacrifice in accuracy. 
All units should, of course, be located by some method that will preclude 
personal choice by the sampler. Randomness within fields is not essential 
in ordinary sampling, since the field itself is the primary sampling unit. 
Number of Field Stops for a County or District 
It has been shown that between-held variation is more important than 
that within held. In previous studies research workers have suggested a 
standard error of 0.125 egg pod per square foot as satisfactory. This level 
of precision is usually easier to reach in a low population than in a high 
one. It will be recalled that in the lightly infested area in South Dakota 
it could be achieved by moderate within-held sampling of 10 helds 
(table 4.). 
Using the data from the 10-county survey of north-central Montana, 
the authors attempted to determine the number of held stops needed for 
each county and for the 10 counties grouped together as a district. The 
held margins were taken into account, being weighted on the basis of 95 
percent for the helds, 5 percent for the margins. The allowable standard 
error of the mean was set at 0.125 egg pod per square foot. This concept 
of allowable standard error as a constant arithmetic hgure is a useful one 
for the level of populations usually encountered. For sparse populations 
it would lead to using only a small number of units (table 8) and some 
minimum should be specihed. 
Table 8. —Number of field stops necessary to survey each of 10 counties in 
Montana, allowing a standard error of 0.125 egg pod per square foot 
County 
Approximate 
population 
mean 
Stops needed to obtain a 
standard error of 0.125 in— 
Fields 
only 
Fields and 
margins 
Judith Basin. 
0.1 
3 
3 
Liberty. 
.4 
10 
9 
Toole. 
.4 
12 
12 
Fergus. 
.4 
13 
12 
Hill.*. 
.4 
28 
27 
Chouteau. 
.5 
21 
20 
Cascade. 
.5 
26 
24 
Blaine. 
.6 
26 
25 
Teton. 
.6 
37 
35 
Pondera. 
.8 
16 
15 
