necessary to obtain information year by year 

 for several years . 



It also seems almost or quite impossible 

 to select any given comparatively short period 

 such as a day, a week, or even a month, and to 

 consider it as representing an entire season. 

 Not only does the fishing change considerably 

 throughout the season but these changes are apt 

 to occur at different calendar dates in different 

 years, in accordance with phenological factors . 



The next idea to consider is to what ex- 

 tent a given part of the area can be used to 

 represent the whole. It has been shown above 

 that for the season under consideration, fairly 

 good mathematical correlation exists between 

 the fishing success of certain localities and 

 pools and that of the entire censused area. It 

 would have been possible, for instance, to have 

 obtained a reasonably accurate figure for the 

 average fishing success of all pools for the sum- 

 mer of 1945 by ascertaining the figure for pool 

 10 alone (or pool 8 alone) and multiplying by the 

 proper factor . Of course the catch is that there 

 was no way of knowing in advance what this prop- 

 er multiplication factor might be. Furthermore, 

 there is no assurance that having obtained this 

 factor for one year it could be used with accuracy 

 for any other year . For any one of a number of 

 obscure reasons, fishing might get relatively 

 better or poorer in one pool than in another 

 from one year to the next. 



Furthermore, as shown above, fishing 

 success from week to week or from day to day 

 throughout a season might have been derived 

 for the entire area, from that for an individual 

 pool or locality. Again, there was no way of 

 knowing in advance what multiplication factor 

 would be needed and no assurance that the same 

 factor would apply in some other year . Also, 

 this procedure would not be without some con- 

 siderable errors. For instance, the week -by- 

 week fishing success in pool 10 in the summer 

 of 1945 showed a good statistical correlation 

 with that for the total of all pools . However, 

 had the former been used to compute the latter, 

 week by week, the calculated figure would have 

 varied from the actual figure by as much as 

 50 percent for certain individual weeks . The 

 same statement holds true for similar calcula- 



tions using the weekly averages for pool 8 . 



There is even greater error in attempt- 

 ing to use any one locality as a representative 

 sample of its pool than in using a pool (or local- 

 ity) to represent the over -all area. There is a 

 wide fluctuation of fishing success for any given 

 period among the various localities in a pool, a 

 part of which is based upon the difference in 

 species composition of the catch between the 

 localities. 



For these waters, therefore, it would 

 appear that accurate averages for fishing suc- 

 cess can be obtained only by making a census 

 every year (at least for a period of several years), 

 throughout the entire season, and for all or most of 

 the pools and the localities within the area. 



There remains, however, the possibility 

 of drastically reducing the effort and cost in- 

 volved in the creel census by randomizing the 

 sample on a daily basis. Figures given above 

 make it appear that reasonably good averages 

 for the season could be expected from a census 

 conducted on one day out of (say) 5 througjiout 

 the season. With good luck, even every tenth 

 day might produce a sufficiently accurate aver- 

 age for the season. Every fifth day would be 

 much safer however . Every seventh day would 

 be a poor choice because of the great variance 

 in numbers of fishermen and fishing success 

 throughout the days of the week . 



For the sake of randomness, the begin- 

 ning day of a fifth -day series should be chosen 

 by lot, especially since each season has several 

 holidays and also certain key fishing days, such 

 as the opening day of the season, the Sunday be- 

 fore Labor Day, etc . Also, once decided upon, 

 the schedule should be adhered to strictly, re- 

 gardless of weather and other fishing conditions. 

 It is possible to set up some sort of system of 

 alternation, whereby the field worker could 

 cover certain localities or portions of his ter- 

 ritory on one series of fifth -days and another 

 portion on another series . 



While yielding the essential information 

 regarding the all-over seasonal average, this 

 system of reduced sampling probably would 

 mean a reduction in the accuracy of tracing 



56 



