Biometrical Letters Vol. 44(1), 2007, pp. 1-15

The objective of the paper is to present a modification of a well-known model with the split plot design by adding new parameters and applying a hierarchical classification as well as a multivariate analysis of variance. The issue is illustrated by the data gathered in the course of a three-year experiment in which changes in a weed community under the tillage systems and doses of herbicides were studied. In the multivariate analysis of variance the weed species were treated as variables and the relative abundance of species as the studied feature. The added parameters describe the effects of the times of observations and the relevant interaction effects. The data were analyzed for each year separately. Herbicides were applied between the first and second observation of the quantity of weeds; thus there were no effects of the herbicide doses for the first time of observation in the first year. Due to this, appropriate restrictions were added to the model. Additionally, in the case of rejection of the null hypothesis the analysis was continued by applying the multiple comparison formula for some chosen sources of variation for the data from the last year of the experiment. Tillage systems, as well as terms of observations and interaction of these factors, caused significant changes in the relative abundance of the weed community. Moreover, in 1999 and 2000 significant differences occurred in doses of herbicides. There were no significant changes observed in the weed community caused by the interaction of tillage systems, terms of observations and doses of herbicides.

herbicide doses, relative abundance index, Roy's test function, simultaneous multivariate multiple comparisons, tillage system.