Sampling at intervals . As described above, all 

 of the creel census here reported upon was per- 

 formed on a day-to-day basis continually 

 throughout the season. It was not possible to 

 visit each locality each day, but a rotation 

 scheme was used so that several individual local- 

 ities within a larger area would be sampled on 

 any given day, and any given locality would be 

 sampled with a certain degree of regularity 

 throughout the season. The aim was that for a 

 larger division such as a pool or for the entire 

 area as a whole, the day-to-day samples would 

 be randomized and representative and would 

 afford fairly complete coverage . 



It is of interest to ascertain to what ex- 

 tent a given portion of the total season's sample, 

 randomized as to time, would be found to be 

 representative and typical of the whole. That is, 

 could a sample consisting of the census on every 

 fifth day, every seventh day, every tenth day, 

 etc., be made to yield satisfactory over-all 

 averages? 



Table 23 presents some of this informa- 

 tion for the summer of 1945 . One of 5 days was 

 chosen by lot. The figures for numbers of fish 

 and numbers of person-hours for this day and 

 every fifth day throu^out the entire season were 

 totaled. From this the average value of fish per 

 hour was computed. This figure amounted to 

 0.48. This is reasonably close to the over -all 

 average of 0.50. As it happened, the figures 

 for every fifth day added up to 23 .5 percent of 

 the person -hours fished for the season rather 

 than the theoretical 20 percent . 



In the same marmer, averages were 

 computed for every tenth day (starting on a day 

 chosen by lot), and for every Sunday . The tenth 

 day total amounted to 9 . percent of the total 

 fishing hours (rather than the theoretical 10 per- 

 cent), witii an average fishing success of 0.51 

 fisn per hour. The Sunday total amounted to 

 31.2 percent of the grand total fishing hours and 

 indicated an average fishing success of 0.43 fish 

 per hour. This last figure is lower than the 

 everyday average, by an amount which probably 

 is significant. 



The idea is pursued further in the compu- 

 tations presented in table 24. Here are given the 



averages based on every fifth day, starting with 

 each of 5 days in sequence; every tenth day 

 starting with each of 10 days in sequence; and 

 every seventh day, using each day of the week . 

 Each of the figures in the column representing 

 the fish per hour is of course to be compared 

 with the 0.50 of the grand total average. For 

 each of the fifth-day sets, and for each of the 

 tenth -day sets, the average fish per hour is 

 reasonably close to 0.50, the range of variation 

 being from 0.45 to 0.57. When talcen one day 

 of the week at a time, the averages are higher 

 than the grand total average for each day of the 

 week except for Sunday and Wednesday, for 

 which they are appreciably lower. It may be 

 conjectured that Sunday fishing actually is done 

 by fishermen who are less experienced and less 

 skilled than those who fish througjiout the week . 

 Many families combine Sunday fishing with pic- 

 nicking and boat riding, and therefore perhaps 

 do not take their fishing as seriously as do the 

 dyed-in-the-wool fishermen who fish on week 

 days. The lowered average for the every-Wed- 

 nesday set of figures possibly can be accounted 

 for by the fact that in 1945 two major holidays, 

 Memorial Day and the Fourth of July, fell on 

 Wednesday . 



Table 25 gives for each of 4 creel census 

 seasons the average catch per hour based upon 

 fifth-day samples. In each instance the particu- 

 lar day out of 5 was chosen by lot and the sample 

 was constructed simply by using the field data 

 which had been recorded on the given days . The 

 last two columns in the table give a comparison 

 of the average fish per hour for the season, as 

 obtained in this manner, with the over -all total 

 average . The figures agree fairly closely for 

 each of the seasons except for the summer of 

 1944. Here the disagreement probably results 

 from the small size of the total sample and 

 hence of the subsample . 



Further comparison of averages based 

 on fifth -day samples, with total sample averages, 

 may be had from table 26 . Here several of the 

 pools and a few separate localities are treated 

 individually. The fifth -day series starts with 

 May 5, the day chosen by lot, as above. With 

 most of these individual pools as with the total 

 (all pools), this particular series of fifth-days 

 amounted to appreciably more than the theoretical 



46 



