Within the last decade new statistical methods have become available in SAS to correctly estimate analysis of variance models containing both fixed and random effects (mixed models; PROC MIXED). These methods offer several advantages over traditional fixed-effect approaches, not the least of which is appropriate parameter estimation for inference spaces that extend beyond the specific fields, plots, or experimental subjects examined. However, a number of errors are commonly made by researchers employing these new methods. Our poster addresses five of these common errors including: 1) differences in the hypotheses of interest between fixed-effect and mixed models, 2) incorporation of appropriate error terms in the RANDOM statement of PROC MIXED, 3) incorporation of appropriate corrections to degrees of freedom, 4) recognition and proper specification of hidden nesting, and 5) use of adjusted mean separations and slices to minimize inflation of type I errors during hypothesis testing.
Keywords: Mixed models, ANOVA
Back to Display Presentations, Section Cf. Quantitative Ecology
Back to Posters
Back to The 2003 ESA Annual Meeting and Exhibition