An evaluation of analysis methods to eliminate the effect of density variation in property comparisons of wood composites

The objective of this research was to evaluate commonly used data analysis methods in property comparisons of wood composites to eliminate the effect of the density variation among board test specimens and to suggest a more r easonable and robust method. The methods reviewed included average, specifi c strength, and analysis of covariance. The indicator variable method was a lso applied to the property comparison and compared to the other methods. T he modulus of rupture of wood fiber/polymer Buff composites manufactured wi th different material combinations and press temperatures was tested in the experiment for evaluation of the different analysis methods. The results o f this study indicated that the statistical analysis method employed was ve ry important in the study of the physical and mechanical properties of wood composites. The specific strength method is limited to the analysis of str ength comparison for the high density composites. The analysis of covarianc e can be applied to all the property comparisons for either high or low den sity composites in eliminating the density variation effect. However error exists in the property comparison using the analysis of covariance method w hen the slopes of the regression lines of property vs. specific gravity (SG ) are different for the different composites being tested. The indicator va riable method is shown to be more reliable than the specific strength and a nalysis of covariance methods because it compares the linear regression lin es of property vs. SG by testing both the intercept and slope based on the data in the whole specific gravity range of test specimens.