Tropical Grasslands (2006) Volume 40, 119–125

Presenting summaries of plant density data that are meaningful to your readers — a case for giving the median as well as the mean

C.K. McDONALD AND R.M. JONES

CSIRO Sustainable Ecosystems, St Lucia, Queensland, Australia

Abstract

The distribution of plant density counts is sometimes highly skewed and presenting only the mean can give a misleading impression of the pasture as a whole, but indications of skewness or dispersion are rarely given in the literature. To evaluate this, 89 data sets from field experiments were analysed to examine the relationship between the mean plant density, the median plant density, and simple measures of skewness and dispersion. In new sowings, the mean seedling density, whether sown by hand in small plots or by machine in larger plots, was consistently close to the median density. In established pastures, the mean:median ratio was <1.5 in approximately half of the 39 analyses undertaken but was more than 2 in about one-quarter of the analyses, i.e., the data were highly skewed. The latter instances commonly occurred when the CV of the individual quadrat entries of any plant count was >150%. When this occurs, to avoid misleading impressions, we consider it important for authors to present the median as well as the mean. If the data are not skewed, the authors should present the CV or SD as well as the mean.

Download full article (87 KB PDF)   

  Return to Contributed Articles