ARS | Publication request: Modeling Biased Fineness and Maturity Results.

Submitted to: Textile Research Journal
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: April 26, 2007
Publication Date: July 1, 2007
Citation: Montalvo Jr, J.G., Von Hoven, T.M. Modeling Biased Fineness and Maturity Results.. Textile Research Journal. 77(7): 495-512.2007

Interpretive Summary: The micronaire of cotton is a combined measure of fiber fineness and maturity. The two quantities are important because yarn made from fine fibers is stronger, and mature fibers absorb dye better. Breeders need accurate results to make important variety decisions. Biased fineness and maturity gives way to incorrect fiber perimeter and wall thickness information. In this study, the theory of biased fineness and maturity measures is introduced. Simple equations are derived to express the bias as a percentage error, relative to the non-biased values. The work goes on to examine several case studies; a method is presented to detect both random error and bias, and produce bias-corrected results. The approach is based on the strong relationship between fineness, maturity and micronaire. In effect, a quality control check is run on each sample in the sample set. Additional analyses are not required. This research may help to validate the fineness and maturity database on US cottons, which could positively impact cotton consumption.

Technical Abstract: In Part I of this series, the classical models study included simulations to explain the variability in coefficients of determination (R2) between fineness and maturity, micronaire and fineness, and micronaire and maturity of cotton. Part II emphasized the derivation and testing of three diagnostic models to enhance the R2 and provide information about the analytical quality (accuracy) of the results. In this paper, the theory of biased fineness and maturity measures is introduced and effects on the relationships with micronaire is simulated. Error functions based on Lord’s micronaire model are derived to express biased results relative to the unbiased values. The simulations include three case studies in order of increasing complexity: bias varies linearly with micronaire, bias is nonlinear with micronaire, and random error is added to the bias. Classical and diagnostic model plots of the simulated biased data are presented in detail to readily determine the scientific merit.