Reader Comments

Post a new comment on this article

Comment on "Climate, Deer, Rodents, and Acorns as Determinants of Variation in Lyme-Disease Risk"

Posted by PLOSBiology on 07 May 2009 at 22:14 GMT

Author: Sarah E. Randolph
Position: Professor
Institution: Department of Zoology, University of Oxford, South Parks Road, Oxford, UK
Additional Authors: Paul J. Johnson, Jorn P.W. Scharlemann
Submitted Date: September 06, 2006
Published Date: September 6, 2006
This comment was originally posted as a “Reader Response” on the publication date indicated above. All Reader Responses are now available as comments.

In a recent issue of PLoS Biology, Ostfeld et al. [1] set out to assess the determinants of variable risk of Lyme disease with a model comparison approach using long-term data on a variety of parameters -- temperature, precipitation, and abundances of deer, mice, chipmunks and acorns. As such, the statistical analyses are the crux of the whole paper and the conclusions. Unfortunately, there are two fundamental flaws in the statistical methods.

The main conclusion is that the primary determinants of temporal variation in Lyme disease risk are acorn, white-footed mice and eastern chipmunk abundance. Climatic variables (degree days and precipitation) only weakly determine density of nymphs (DON), nymphal infection prevalence (NIP) and density of infected nymphs (DIN), and there is no support for deer having an impact on tick abundance or infection prevalence. However, not all potential predictors entered into the statistical model were measured independently for each replicate plot used in the experimental design. All climatic variables and direct estimates of deer abundance were recorded not for each of the six replicate plots separately, but were attributes of the entire study site. While climatic variables and deer abundance did not vary among plots on the same sample date, they appear nevertheless to have been entered into the statistical models as characteristics of individual plots. The six plots are sub-samples within the study site, and thus do not provide replication for variables which are constant at the site at the same time point, only pseudo-replication [e.g. 2]. To correct this, we suggest re-analysis, pooling data from all plots for each year to test for the effects of changes in climate and deer abundance over time.

Further, the models were run using individual "plot-years" as observations (58 for climatic and deer variables available over all years for all plots, and 42 for rodent and acorn variables available over fewer years for some plots), thereby confounding variation in time and space. The effect of correlation of observations made on the same plots needs to be incorporated by including plot and year as categoric factors in the statistical models. This might also throw light on the origin of the extreme outlier in Fig. 4 of [1], which is entirely responsible for the relationships as presented.

In summary, conclusions based on these flawed statistical procedures cannot be accepted as reliable until the data are re-analysed using standard practices that do not include pseudo-replication. The contribution of spatial and temporal variation must be clearly separated, both visually in the graphs and statistically in the models. Only then can firm conclusions be reached as to whether climatic variables and deer, or rodents and acorns, or a combination of these, are important determinants of Lyme disease risk.

1. Ostfeld RS, Canham CD, Oggenfuss K, Winchcombe RJ, Keesing F (2006) Climate, Deer, Rodents, and Acorns as Determinants of Variation in Lyme- Disease Risk. PLoS Biol 4(6): e145.
2. Quinn GP, Keough MJ (2002) Experimental Design and Data Analysis for Biologists. Cambridge: Cambridge University Press. 537 p.

Competing interests declared: We declare that we have no competing interests.