Analysis of local variation in plant size.

Attempts to quantify local competitive interactions by examining size and growth of neighboring plants may suffer from several statistical difficulties: 1) when individuals are analyzed both as focal plants and as neighbors of other individuals, observations are not independent and tests of statistical significance may be invalid; 2) effects of competition and local site quality may be confounded. Path analysis and randomization tests may permit more accurate estimation and testing of influences on individual plant growth. Analysis of results of experiments with Impatiens capensis showed that conventional statistical tests may yield serious errors in estimates of significance, and that these errors are of unpredictable magnitude and direction. It is essential to measure or estimate the effects of site quality on variation in plant size. -from Author

Duke Scholars

Author Thomas Mitchell-Olds Biology