Are Currently Available Statistical Methods Adequate for Long-Term Studies?

Abstract

Within the range of approaches available for investigating long-term phenomena, it is found that those which may be suitable for large spatial and long temporal scales often have the least potential for causal inference, while the textbook-style procedures of great inferential power may not yield insights applicable to many scales of interest. Furthermore, methods do exist for the design and analysis of studies across the continuum of potential study types, yet the graduate curriculum has been inadequate in preparing ecologists in these methods, and consequently researchers are often unaware of potentially useful approaches. The discussion group on statistical methods considered the special needs of long-term studies by identifying how long-term phenomena were treated within the range of potential approaches and by evaluating whether long-term studies were unique in design or analytical requirements. Study design was the main focus of the group, though some suggestions were made of specific underused, though potentially useful, statistical methods. Issues not considered for lack of time were ways to affect appropriate matches between study designs/approaches and specific classes of hypotheses, systems, or processes, and a characterization of desirable aspects of good parameters or indices of system response.

Discussion Group 5: C.D. Canham (leader), S.R. Carpenter, S.H. Huribert, W.K. Laurenroth, J.M. Melillo, G.G. Parker (reporter), L.R. Taylor, and R.G. Wiegert.