FISHERY BULLETIN: VOL. 71, NO. 3 



each item was calculated. These average weights 

 were placed in an array in item code order. 

 The records which were defective in purchase 

 weight were then read sequentially, and the 

 item code in each record was used as a sub- 

 script to access the proper average weight 



figure in the array. The average weight figure 

 thus located was inserted into the defective 

 record, completing the recovery process. 



A comparison of results obtained on different 

 types of runs using both data bases shows 

 little difference (Appendix Table 1). 



Appendix Table 1. 



-Comparison of data bases using four indices (all figures in 

 jug/day).' 



' Intake levels in figlday exceeded only by 0.1°o and 1.0% of the test group. 



- DBl — using Data Base 1. BD2 — using Data Base 2. These figures were obtained 

 using preliminary mercury data and may differ slightly from the final results, but the 

 comparative data will not be significantly different. 



626 



