Sample size and area: implications based on long-term monitoring of partridges.

Abstract

The choice of a suitable sample size and size of study area is a serious problem facing experimental research workers, and is often not given the attention it deserves. The difficulty lies in balancing statistical requirements with the logistics of the experiment. With too small a sample or study area, toxic effects of a chemical may remain statistically undetected because of random noise (Type II error). On the other hand, large sample sizes and large study areas which are certain to satisfy statistical needs may be prohibitively time-consuming and expensive. In testing for toxicity effects of pesticides on vertebrates, the problem is often made more acute by the fact that an effect may be indirect rather than direct. For example, newly-hatched young of many bird species are fed on insects (e.g. O'Connor, 1984), and a pesticide-induced decline in insects will have repercussions upon chick survival. Such indirect effects are less easy to detect than direct poisoning, because they are more easily swamped by environmental noise. Poisoning for instance may usually be ascertained by tissue analysis, but starvation may result from a number of different causes. To design experiments in these conditions, prior knowledge is often essential. It may be obtained through a preliminary trial, from experience or from parallels with other similar existing work.
The Game Conservancy, which produces recommendations aimed at integrating modern farming with game and wildlife conservation, constantly faces the problems outlined above. Based on 21 years of ongoing work on the Grey Partridge Perdix perdix in Sussex, this paper draws conclusions concerning sample size and size of study area which are of particular relevance to the question of examining toxicity effects of pesticides in farmland vertebrates.