L.3.3 Bioaccumulation from field data

The power of a comparison can be determined by:

t 1-β =

- t 1-α ,v

(L-32)

When variances are not significantly different, s is replaced by (MSE)2 and v

= N - k df. Using MSE = 0.003763 as above, the power to detect a 10 percent

decrease in mean bioaccumulation below the action level is 0.16, and power to

detect a 50 percent decrease is 0.96. Power for 10, 20, 30, 40 and 50 percent

decreases are given in the output for SAS program BIOACC (Section L.4.2.2).

Less than detection limit data. Recommendations for analysis of

bioaccumulation data with less than detection limit values were developed to

facilitate comparisons of two or more samples. When a sample of contaminant

bioaccumulation concentrations must be compared with an action level or

standard, accurate estimation of the sample mean and standard deviation is

important. In general, this may require different censored data methods than

does the comparison of samples in the previous section. Most recommendations

for censored data methods in estimation problems have been based on relatively

large sample sizes (n = 10 or more). Gleit (1985) identified certain methods that

perform better than others for estimating the mean and variance of normal

populations based on samples of n = 5. The best methods, depending on mean,

coefficient of variation, and amount of censoring, included substitution of DL,

DL/2, or zero, and an iterative method using expected values of order statistics.

The latter method (which Gleit recommended), along with several others

including regression and some maximum likelihood techniques, are available in

UNCENSOR (Newman and Dixon 1990).

Recommendations for censored data methods for estimating mean and

standard deviation when n is small are provided by Clarke and Brandon (1996).

If zero is substituted for all nondetects and the sample mean is greater than or

equal to the applicable action level, then clearly no statistical testing is required

as the mean contaminant concentration cannot be less than the action level.

L.3.3 Bioaccumulation from field data

A field bioaccumulation test may be designed to show differences, if any,

between organisms living at the proposed disposal site and the same species

living in the reference area. Ttissue concentrations in organisms collected from

replicate samples at the disposal site(s) are compared with tissue concentrations

in organisms collected from replicate samples at the reference area, using the

decision tree steps in Figures L-4A and 4B. If comparisons involve organisms

from only one disposal site, then the appropriate statistical comparison

procedures, depending on the results of the tests of assumptions, are the two-

sample t-test for equal or unequal variances, or the t-test for unequal variances

using rankits or ranks (Section L.2.1.1.1).

L39

Appendix L Selected Resource Documents