upon receipt of data from Omni. The values reported for 
each element were followed from shipment to shipment 
to detect any abnormal trends. Three consecutive high 
or low values were taken to indicate the existence of an 
analytical problem. To avoid having all samples in a 
group analyzed during a period of abnormally high or low 
values, whenever possible no more than two samples 
from the same group were sent in the same shipment. 
Bar graphs of control sample analytical values were 
constructed to assist in the evaluation and were sent 
periodically to Omni. 
As a method of checking analytical reproducibility 
over time, in addition to using controls, a minimum of 
20% of the samples from each group was to be dup- 
licated in later shipments; overall duplication was about 
50%. Single analyses were run on the remainder. Dup- 
lication was also used to confirm or reject suspect data. 
Frequent contacts were made with Omni to review the 
status of contract work, including methods of analysis, 
variations in analytical values, their possible causes, and 
recommendations for correction. 
Data Management 
Collection and physical composition data were as- 
sembled from information furnished on the Fish Data 
Label (Appendix Figure 2) returned by collectors with 
each fish. Information that the collector failed to provide 
was added at College Park (e.g., length and weight of a 
whole fish), or gathered from a phone call or letter, when 
possible. Not all the information requested was available 
for all fishes. Depth of catch and age were seldom 
available; sex was sometimes available; latitude and 
longitude of catch were available for approximately half 
the fishes; date of catch, fish length and weight, and the 
number of fish were usually available; and some descrip- 
tion of the catch location was almost always available. 
When only the weight of a headed, gutted, or headed and 
gutted fish was provided, that weight was converted to 
an estimated round weight. 
The collection and physical composition data for each 
group of samples were then entered on a “‘history”’ sheet. 
If the collector did not send enough fish for 10 samples, as 
requested, samples composed of fish from similar dates 
of catch, location, and/or size were grouped together on 
one “history” sheet. These groups represent pages in the 
“history” sheet files and were a necessary means of 
checking the computer tapes. 
When the analytical data were received from the con- 
tractor, they were entered on a coding form (Appendix 
Figure 3) with the data from the “history” sheet and were 
keypunched and entered into the computer. For the lat- 
ter half of the samples, the contractor entered the analy- 
tical data directly on the coding forms. 
To check that the data were entered correctly, a cycle 
of: 1) generate computer printout, 2) check printout for 
accuracy, 3) correct errors and add new data was 
repeated several times. Once all the data were in the 
computer, each group was reviewed using Grubbs’ out- 
lier test (paragraph 4.1 of Grubbs 1969) to determine if 
any values deviated from the group mean more than 
could be anticipated with normal statistical theory to 
remove grossly unreasonable lone values as well as to 
identify periods when values were repeatedly un- 
reasonable. Using this approach, about 1% of the analy- 
tical values was determined to be too questionable for 
use in this report. 
The Microconstituent Data Bank, in which the 
Resource Survey data are stored, is a card-to-tape sys- 
tem with a format similar to the coding form (Appendix 
Figure 3). It was established in 1973 for the purpose of 
storage, retrieval, and tabulation of data generated by 
the Resource Survey and similar studies. Twelve 
program runs, most with several choices of data break- 
down, can be generated with programs presently in the 
system; Table 5, page 119, is an example of one of these, 
the MF17 program run. Other programs list species, 
broken down according to location, and provide a vary- 
ing depth of detail from 1 to 15 elements. They also 
provide differing aggregates of history information. 
RESULTS 
Overview 
The Resource Survey provides information on the oc- 
currence of 15 elements in the fishery resource. It covers 
204 species taken from 198 sites in 7 areas of U.S. coastal 
waters. Excluding controls, a total of 9,347 unique 
samples of 13 different tissues or product forms were 
analyzed. 
The contractor returned approximately 220,000 analy- 
tical values of element levels, including replicates but ex- 
cluding controls. Of these, about 2,900 values were re- 
jected according to the criteria established for data 
checking. Approximately 37,500 analyses were reported 
by the contractor as “‘not detected”’ or ‘‘zero” without 
reference to detection limits and consequently were not 
used to determine mean trace element levels. 
In a project of this magnitude, invariably some infor- 
mation is lost and a degree of uncertainty introduced into 
the data as a result of logistical and analytical difficul- 
ties encountered during various phases of the work. 
There were periods during this survey when analytical 
results reported by the contractor reflected problems of 
this kind. At times, imprecision in the data could be as- 
sociated with changes in methodology and analytical 
procedures. During some periods, many of the samples 
appeared to have been contaminated with one or more 
elements. At other times, only occasional unreasonably 
high values were reported. 
When checking data, it is relatively simple to find a 
single deviant value among the results of many analyses 
on the same sample or even the same species, i.e., when 
comparing similar fish. It is generally impossible to dis- 
tinguish between deviant and acceptable values, 
however, when few analyses are available. Therefore, 
there is a much higher chance of reporting an inaccurate 
value of the mean element content of a species 
