Statistical advantages in, and characteristics of, data from long-term research
Richard A. Fleming · The Forestry Chronicle · 1999
Scientifically, long-term research is the best approach for investigating phenomena involving slow, subtle, and long-period cyclic change. The importance of such phenomena is emphasized by recent work in ecological theory which suggests that they have a constraining, controlling influence on other faster ecosystem processes. Four short-term "alternative" approaches (retrospective analysis, fast system analogues, simulation models, and space-for-time substitution) each have drawbacks compared to long-term research in studying these kinds of phenomena. Space-for-time replication, however, could complement long-term research well. Various statistical concerns (missing data, autocorrelation, statistical power, precision, bias, and spurious correlation) are briefly discussed in the context of justification for, and recommendations during the conduct of, long term research. Key words: statistics, long term research, data analysis, statistical power, space-for-time substitution