Comparison of models of fixed and mixed effects on the analysis of an experiment with mutant strains of cellulotic fungus Trichoderma viride
Abstract
Two models of fixed effect and one mixed were used to compare models of fixed and mixed effects on the analysis of measurements repeated in time through an experiment with mutant strains of cellulotic fungus Trichoderma viride. The data processing was conducted with the statistical software Infostat Version 1 and SAS (2007) version 9.1.3. Multivariate analysis of variance (MANOVA) was conducted from the
results of the Pearson correlation matrix, an univariate analysis of variance (ANOVA), according to divided plot design and an analysis of variance using the mixed model (PROC MIXED of SAS with option Repeated). The strains and the sampling times were considered as fixed effects, and the experimental units (Erlenmeyer flasks) were considered as random effect (subjects). The selection criteria (Verosimilarity,
Akaike and Bayesiano) were taking into consideration for a better model adjustment. The interaction between sampling times and treatments (strains) was significant for P < 0.05 in both methods used. The results showed that the analysis with mixed models was more success, because it solves the failing of the basis hypothesis, solves the limitation of the multivariate analysis of variance and gives higher flexibility
and information when selecting the model of better fit. It also allows analyzing data bases in unbalanced designs.
Key words: repeated measurements, univariate analysis of variance, mixed model.
results of the Pearson correlation matrix, an univariate analysis of variance (ANOVA), according to divided plot design and an analysis of variance using the mixed model (PROC MIXED of SAS with option Repeated). The strains and the sampling times were considered as fixed effects, and the experimental units (Erlenmeyer flasks) were considered as random effect (subjects). The selection criteria (Verosimilarity,
Akaike and Bayesiano) were taking into consideration for a better model adjustment. The interaction between sampling times and treatments (strains) was significant for P < 0.05 in both methods used. The results showed that the analysis with mixed models was more success, because it solves the failing of the basis hypothesis, solves the limitation of the multivariate analysis of variance and gives higher flexibility
and information when selecting the model of better fit. It also allows analyzing data bases in unbalanced designs.
Key words: repeated measurements, univariate analysis of variance, mixed model.
Full Text:
PDF (Español (España))Refbacks
- There are currently no refbacks.