Electronic Scientific Paper Archive

Designing Experiments that Control for Spatial and Temporal Variation

Peter Petraitis
Department of Biology, University of Pennsylvania, Philadelphia, PA 19104-6018, USA e-mail: ppetrait@sas.upenn.edu

Abstract

Spatial and temporal variation can cause problems in designing and conducting experiments. An introduction to methods for controlling spatial and temporal variation in ecological experiments is provided in this article. Failure to consider spatial and temporal variation often causes researchers to lay out experiments incorrectly. The challenge is to design experiments that not only reflect the natural variation seen in the field but also control for the variation so that statistical tests have sufficient power. Spatial variation is usually controlled by grouping observations and treatments into blocks. Blocks can be laid out in a number of ways and Analysis of Variance (ANOVA) approaches to control for block effects are discussed.
The control of temporal variation presents special difficulties because data are often serially correlated and so observations are not independent. Use of intervention analysis and repeated measures analysis of variance to control for temporal variation are discussed. Ecologists have also used experimental designs which are known as BACI designs (i.e., Before-After-Control-Impact design) and can be extended to include multiple control and/or impact sites. Intervention analysis, BACI designs, and their extensions have subtle differences because of different assumptions about not only temporal variation but also spatial variation. Several recommendations are given. These include: 1) the need to have good statistical advice before starting an experiment, 2) the need to have a sufficient number of replicates that are spread over the range of spatial and temporal variation, and 3) the need to correctly control for serial correlation.

Keyword: spatial variation,temporal,experimental design,ANOVA

Introduction

Ecological data derived from experiments and observational studies done in the field are very often variable because patterns in nature are masked by temporal and spatial variation in physical and biological factors. Even though ecologists are well aware of the problems caused by natural variation, the ecological literature is filled experiments that either ignore the difficulties caused by spatial and temporal variation or deal with them in an inappropriate manner. Here I provide an introduction to methods for controlling spatial and temporal variation. There is a vast literature in this area and I can only offer some guidelines and provide a list of helpful references.

Conclusion

In closing, I would like to offer several suggestions. While my advice is framed within the context of the problems raised by spatial and temporal variation, these are common-sense notions that are applicable to any field experiment.
1. Seek advice about experimental design and statistical analysis before you start an experiment, not after you collect the data.
2. Spread your replicates over the ranges of spatial and temporal variation that are of interest. Avoid setting up an experiment in which all the

Reference

  1. Hurlbert, S.H. 1984. Pseudoreplication and the design of ecological field experiments. Ecological Monographs 54:187-211. Analysis of variance
  2. Potvin, C. 2001. ANOVA: experimental layout and analysis. In S.M. Scheiner & J. Gurevitch (ed.): Design and Analysis of Ecological Experiments, second edition. Oxford University Press, Oxford, UK. pp. 63-76.