The analysis of data from a BACI study
by
K.G. Russell
Institute for Mathematical Modelling and Computational Systems, University of Wollongong, N.S.W. 2522, AUSTRALIA

BACI (Before After/Control Impact) studies are commonly used to assess the impact of some major occurrence on the immediate environment with reference to several control sites. Usually there is only one impact site available for monitoring before and after the event in question.

In the investigation described here, the event under consideration was a residential development near Wollongong, south of Sydney. The aim of the BACI study was to assess the impact of the development on nearby waterways. Response variables included the observed numbers of various species of invertebrates that habitually occupy these waterways. Sampling took place at three locations on each of five creeks near the development, for periods of two years before and after the development took place. Some locations were considered to be unaffected by the development, while other sites could potentially be affected. It was sometimes impossible to collect water samples due to a lack of water in the creeks.

This paper will describe two forms of statistical analysis that have been used. The first is multi-level modelling. The second involves a regression analysis of a reduced data set formed from contrasts within the full data set. Attention will be restricted to just two of the many response variables that are available.

(The data were supplied by Glenn Johnstone, of the Department of Biological Sciences, University of Wollongong. The speaker has greatly benefited from discussions with Kathy Haskard, of BiometricsSA, Adelaide, South Australia. However, any faults that remain are the responsibility of the speaker.)

Date received: August 29, 2001