Variability of these data is naturally high and is further complicated by 
less than ideal stratification. Table 10 gives the cumulative kill on 
private land only, and by periods, for both the ground survey and the mailed 
survey. 
Table 11 - COMPARISON OF HARVEST ESTIMATES BY PERIODS 
(Private Land Only) 
Period Horicon Plan .05 Limits Mailed Survey -05 Limits 
Oct. 7-9 2,510 + 134 1,872 + 43% 
Oct. 10-14 2,700 + 12% 3,288 + 30% 
Oct. 15-16 1,005 + 124 1,616 + 10h 
Total Sample 6,215 + &f 6,776 + 13% 
Total Known 4 340 4340 
Grand Total 10,555 11,116 

Table 11 presents the estimated harvest for each of the three periods 
of the mailed questionnaire, and data from corresponding periods for Horicon 
ground survey. The 0.05 limits for the estimates from the samplings are also 
given. The variance of estimates of the kill on private land is relatively 
high. However, this variability is applicable to only about 60 percent of 
the total kill each year. In terms of the total kill, then, it is not so 
excessive. The sampling intensity can be reduced somewhat and still provide 
reasonable precision in terms of the total kill when a portion of the total 
is known from registration, rather than estimated from a sample. 
This variability may possibly be reduced in the future by improvement 
in the stratification and in allocation of the sample proportionally to 
variance within each stratum. 
Figure 4 represents the distribution of the goose harvest around Horicon 
Marsh and the distribution of goose hunting effort. 
The Return 
The return from the 1960 questionnaire (table 12) indicates that a pre- . 
liminary estimate, not corrected for nonresponse, could be made in about 
5 days following the end of the period covered by the questionnaire. For all 
practical purposes the return was complete about le days from the end of the 
period, and a final corrected estimate could be made as soon as the required 
number of nonrespondents had been interviewed. The correction for nonresponse 
reduced the estimates by an average of 38.5 percent. This correction may 
differ with the goose population or with hunting conditions. The nonresponse 
check is a necessity at present. After a few years of experience it may be 
possible to estimate the progress of the harvest relative to previous years 
by comparing the uncorrected estimates, but for the present more information 
is required. 
ry), 
