Statistical design, analysis, and presentation issues

Abstract

Presentation issues provide a perspective from which to comment on the role of statistics in applied biological research. Full details are required of the experimental layout, and of reasons for the choice of design. Gains from effective design may include: improved precision and clearcut results; reduced cost and/or effort; information that a less satisfactory design could not have provided; and clues that will help in planning further experiments. Failure to distinguish distinct strata of variation is a common source of error, both for design and for analysis. Valid treatment comparisons require replication at the level of whole treatment units. Points that require attention in analysing data and reporting results include: limitations in the design and in the data; evidence of appropriate checks; the tying of statistical evidence to biologically meaningful questions; translation of results back into subject matter terms; and lucid presentation. Analyses must reflect the structure of the data; thus a trend or response surface analysis is appropriate when levels of one or more factors vary quantitatively. Attention is drawn to some major recent advances in approaches to design and analysis.

Keywords:

• biometrics
• statistics
• presentation
• graphics