Cotton genotype effect on fiber quality distributions
In the 2004 crop year, 151 genotypes of Upland cotton and 10 genotypes of Pima cotton were planted in the United States. Many previous studies showed that cotton genotype is the major contributor to overall quality variation. Many other studies demonstrated the importance of distributions of fiber quality parameters in predicting end product quality. However, much of the previous research conducted on fiber quality parameters used either commercial bales or cotton bolls. Due to process effects or small sample size, neither of these truly and fully reflects the genotype effects on distributions of fiber characteristics. This study will fill the gap in the existing literature by examining the relationship between cotton genotype and the distribution of fiber quality parameters. To minimize all other effects, different genotypes of California San Joaquin Valley cotton were planted on two locations with two field replicates, and ginned using the same process at the same laboratory. Samples that are large enough to represent the cotton genotype were used. The cotton quality parameters were limited to length by weight, length by number, maturity, and fineness because these are shown to be important factors of yarn quality. Test equipment was limited to the AFIS (Advanced Fiber Information System). Histogram data from the AFIS were analyzed using several different statistical methods to determine whether or not the distributions were genotype dependent. To compare the shape only, standardized data was used. When the Kruskal-Wallis test was applied on whole distribution, genotype does not affect fiber quality distributions. Cluster analysis revealed in certain cases (33%~90%), the genotype affect is greater than replication variances. When the Kruskal-Wallis test was applied on part of the distributions, genotype was shown to affect fiber quality distributions. Ranges close to standardized mean values (between 0 and -1.0 sigma) were always affected for all fiber quality parameter in all years. Mathematical models that described the distribution of each genotype were built for each fiber quality parameter. Out of 352 data sets, 339 data sets showed R2 values of greater than 0.98 and the remaining greater than 0.965. The Kruskal-Willis test and cluster analysis were used without standardization as well and revealed that genotype affects fiber quality distributions at all fiber quality parameter in all years. Pair wise comparisons also enabled us to investigate which genotype is statistically different against which genotype was found. Cluster analysis also detected a genotype effect.