Statistical tests and power analysis for three in-vivo bioassays to determine the quality of marine sediments.
Nelly van der Hoeven1*, Belinda J. Kater2§ and Jan F. Pieters2.
§: Present address: RIVO - Centre for Shellfish Research, P.O. box 77, 4400 AB Yerseke, The Netherlands
ECOSTAT, Vondellaan 23, 2332 AA, Leiden, The Netherlands
- National Institute for Coastal and Marine Management, Field station, Jacobaweg 2, 4493 MX Kamperland, The Netherlands
* Corresponding Author:
Nelly van der Hoeven,
2332 AA Leiden,
tel.: (+)31-(0)71-5315011; fax: (+)31-(0)842-116988;
Statistical tests are recommended for three marine sediment in-vivo bioassays. In two bioassays (Corophium volutator and Echinocardium cordatum), the mortality in the sediment is compared with that in a control. An unconditional 2×2 test is recommended. For one bioassay (Rotoxkit M with Brachionus plicatilis), mortality in a dilution
series of pore water is compared with the mortality in a control. The Williams' test for trends is recommended.
For each of these tests the power to assess an effect has been calculated. The number of replicates recommended in the standardized test protocol only allow to observe large effects in almost all (95%) of the experiments. Given the control mortality rates estimated from a large set of controls, a power of 95% will only be reached if the mortality
rate in the tested sediment is over 30% for C. volutator and almost 60% for E. cordatum. To reach this power for bioassays with B. plicatilis, where 5 concentrations are compared with a control, the mortality rate in the Lowest Effect Concentration should be about 35%.
As alternative to no effect testing, it is suggested to test whether the effect of a treatment remains below some chosen minimal relevant effect (MRE). Given an MRE at a fixed mortality rate of 25% and α=0.05, at least 55 individuals are necessary to be reasonable sure (95%) that a mortality of 10% will not be declared toxic incorrectly.
The tests for mortality are based on the assumption that survival of individuals within a test vessel are independent. We have described a method to test this assumption and applied it to the data on C. volutator.
keywords: NOEC; ECx; Williams' test; Cochran-Armitage Trend test; Unconditional 2×2 test
Environmetrics 13: 281-293, 2002
Statistical design of biological research
Development of statistics and experimental design for several laboratory ecotoxicological bioassays (with the crustacean Corophium, the sea-urchin Echinocardium, the rotifer Brachionus and larvae of the oyster Crassostrea, among others. One of these studies is described in van der Hoeven et al., 2002
- Investigation into the number of replicates needed for the desired power in a field experiment with a herbicide.
- Investigation in the number of replicates and sub-samples needed in an ecotoxicological and histopathological study with fish (Pimephales) and snails (Marisa).