L.2 Statistical Methods for Water Column Tests

complete blocks design should be used when the experimental units are placed

on or in several different tables, benches or water baths (i.e., "blocks"). Each

block holds a certain proportion of the experimental units. Treatments are

assigned to experimental units randomly within each block, and each block

contains an equal number of replicates of each treatment. Either of these designs

is acceptable, providing the principles of replication, randomization,

interspersion, and controls are followed. Adherence to the principles of

experimental design ensures that the most basic assumption of statistical

hypothesis testing, the assumption that treatments are sampled independently, is

met.

L.2 Statistical Methods for Water Column Tests

L.2.1 Water column toxicity tests

The objective of the analysis of water column toxicity test data is to assess

the evidence for reduced survival because of the toxicity of suspended plus

dissolved dredged material constituents. If reduced survival is evident, then the

median lethal concentration (LC50) or effective sublethal concentration (EC50) of

the dredged material is calculated from a serial dilution experiment. Figures L-1

and L-2 provide an overview of water column toxicity test data analysis. Control

survival must be $90 percent or some other appropriate value, otherwise the test

must be repeated. At the end of the exposure period, the effects, if any, on the

survival of the test organisms should be clearly manifest in the 100 percent elu-

triate concentration. When the dilutions are prepared with other than control

water, the dilution water treatment is preferred over the control water for the data

analysis. If the elutriate survival exceeds the control survival, then the toxicity

test indicates no adverse impact from the dredged material.

L.2.1.1 Comparison of 100 percent elutriate and dilution water

L.2.1.1.1 Methods

Two-sample t-test. The usual statistical test for comparing two independent

samples, such as the 100 percent elutriate and the dilution water in water column

toxicity tests, is the two-sample t-test (Snedecor and Cochran 1989). The

t-statistic for testing the equality of means 1 and 2 from two independent

samples with n1 and n2 replicates is:

t = ( x1 - x 2 ) /

s pooled (1/ n1 + 1/ n2 )

(L-3)

where s2ooled, the pooled variance, is calculated as:

s pooled = [ s1( n1 - 1) + s 2( n2 - 1)] / ( n1 + n 2 - 2)

(L-4)

and where s2 and s2 are the sample variances of the two groups. If the sample

sizes are equal (n1 = n2), then:

Appendix L

Statistical Methods