Table 7 



Tests for Estimating Sampling Reliability of 1950 C lamshell 

 Bucket Samples Computed Using Values Transformed by /x + 3/8 



Bullraked Area Control Area Dredr;ed Area 

 Statistics 1st 50 2nd 50 1st 50 2nd 50 1st 50 2nd 50 



Arithmetic Mean 2.2U 

 Standard Deviation 1.00 

 Standard error of mean O.lU 



Standard error of 



difference 

 between means 



Normal deviate 



1/ This deviate would be exceeded by 100^ of trials. Therefore differ- 

 "" ence is due to chance and sampling is considered reliable. 



2/ This deviate would be exceeded by 50-55$^ of trials. Therefore dif- 

 ference is due to chance and sampling is considered reliable. 



3/ This deviate would be exceeded by 90% of trials. Therefore differ- 

 ~ ence is due to chance and sampling is considered reliable. 



In each case the normal deviates are so small that variations of 

 this magnitude would be expected in repeated sampling of a single popula- 

 tion. Since we already know that each pair of series were taken from 

 one plot with a single population these tests show that series of $0 

 samples each provided reliable estimates of the population. 



The difference between the mean of the 19U9 survey in the control 

 area was compared .with the means of the two 1950 series in this unfished 

 plot. This decrease from 11.86 to 7.03 clams per sample as indicated by 

 the first sampling series of 1950 would be expected to occur by chance 

 only one time in one thousand. If the value of 7-71 clams per sample 

 indicated by the second series in 1950 is used, the probability becomes 

 once in one hundred. Therefore this decrease in abundance of clams was 

 real, and was not attributable to sampling error. Reasons for the dis- 

 appearance of clams in the control area are given in the text. 



Uo 



