2017 ESA Annual Meeting (August 6 -- 11)

COS 95-1 - Log(arithms) are confusing, but we consume them like a band of beavers: Findings from a survey of ecologists and a meta-analysis

Wednesday, August 9, 2017: 8:00 AM
C122, Oregon Convention Center
Duncan Menge1, Anna C. MacPherson2, Thomas A. Bytnerowicz1, Andrew W Quebbeman1, Naomi B. Schwartz1, Benton Taylor3 and Amelia Wolf1, (1)Ecology, Evolution, and Environmental Biology, Columbia University, New York, NY, (2)American Museum of Natural History, New York, NY, (3)Organismic and Evolutionary Biology, Harvard University, Cambridge, MA
Background/Question/Methods

Ecologists commonly present data in logarithm-scaled space. Based on our experiences, we hypothesized that data presented in logarithm-scaled space, and particularly with logarithmic values (e.g., -1, 0, 1, instead of 0.1, 1, 10), would mislead readers about the relationships between variables. To test this hypothesis we developed an online survey and, with the help of the ESA office, distributed it to the ESA membership. The survey asked respondents to identify four pieces of information about relationships between two variables, using the example of rabbit and chipmunk populations as a function of distance from the edge of a habitat. Each respondent saw figures and tables from randomly selected datasets and randomly selected scale formats (linear scale and values, logarithmic scale and linear values, or logarithmic scale and values). The four pieces of information were (1) whether a population increased or decreased, (2) which population changed more steeply, (3) the sign (positive, zero, or negative) of the population size at the edge of the habitat, and (4) whether populations changed in an accelerating vs. constant vs. decelerating manner. Over 1000 ecologists responded to the survey.

Results/Conclusions

Most survey respondents (98-99%) readily identified whether populations were increasing or decreasing with distance from habitat edge for linear and logarithmic scales. By contrast, respondents identified other aspects of the relationship much more readily on a linear vs. logarithmic scale (86% vs. 32% for which population changed more steeply, 92% vs. 43% for population size at the habitat edge, and 91% vs. 30% for accelerating vs. constant vs. decelerating). Information was clearer to respondents in figures than in tables when presented on a linear scale. For information presented on a logarithmic scale, figures were clearer than tables for some pieces of information, but tables were clearer for other pieces of information. Respondents also stated a strong preference for viewing information: 96% preferred a linear scale, and 84% preferred linear values for data presented on a logarithmic scale. Based on our results, we recommend that ecologists present data on a linear scale and with linear values unless there is a compelling reason for presenting logarithmic scales or values. In cases where there are compelling reasons to use a logarithmic scale or values, we recommend that authors guide readers more carefully through their display items to avoid misunderstanding.