The mismatch between current statistical practice and doctoral training in ecology

Abstract

Citation: Touchon, J. C., and M. W. McCoy. 2016. The mismatch between current statistical practice and doctoral training in ecology. Ecosphere 7(8):e01394. 10.1002/ecs2.1394Abstract. Ecologists are studying increasingly complex and important issues such as climate change and ecosystem services. These topics often involve large data sets and the application of complicated quantitative models. We evaluated changes in statistics used by ecologists by searching nearly 20,000 published articles in ecology from 1990 to 2013. We found that there has been a rise in sophisticated and computationally intensive statistical techniques such as mixed effects models and Bayesian statistics and a decline in reliance on approaches such as ANOVA or t tests. Similarly, ecologists have shifted away from software such as SAS and SPSS to the open source program R. We also searched the published curricula and syllabi of 154 doctoral programs in the United States and found that despite obvious changes in the statistical practices of ecologists, more than one-third of doctoral programs showed no record of required or optional statistics classes. Approximately one-quarter of programs did require a statistics course, but most of those did not cover contemporary statistical philosophy or advanced techniques. Only one-third of doctoral programs surveyed even listed an optional course that teaches some aspect of contemporary statistics. We call for graduate programs to lead the charge in improving training of future ecologists with skills needed to address and understand the ecological challenges facing humanity.

Trends in ecology: shifts in ecological research themes over the past four decades

“…Modeling, a tool applied in nearly every subdiscipline, has the highest frequency and is at the center of the thematic space. The dominance of modeling as the most prevalent concept theme may reflect recent advances in the complexity and diversity of statistical tools used to analyze newly available, complex datasets (Touchon and McCoy ; Courchamp and Bradshaw ). Community structure and community processes are both centrally located in the thematic space, which reflects that communities occur as ecological units across subdisciplines.…”

Evaluating the popularity of R in ecology

“…First, R has an enormous amount of resources. Ecological data and questions are becoming increasingly sophisticated and as a result ecology has become an increasingly quantitative, statistical field (Higgs , Touchon and McCoy ). Through CRAN and over 13,000 packages, R offers almost all existing statistical models for ecological analysis (R CRAN Task Views: Environmetrics; Phylogenetics; Spatial).…”

Could ecologists be more random? Straightforward alternatives to haphazard spatial sampling

“…Furthermore, it is now much easier to manipulate spatial data without the use of specialised GIS software. The GIS functionality available in R (R Core Team), an opensource statistical programming language and platform used widely by ecologists (Touchon and McCoy 2016), has greatly improved in recent years. This is largely due to the work of Roger Bivand and colleagues in developing the 'sp' and 'rgdal' packages, which contain many of the functions provided by specialised GIS software (Bivand 2006, Bivand et al 2008).…”