Type of Document Dissertation Author Zhou, Marvellous Mabeza Author's Email Address firstname.lastname@example.org, email@example.com URN etd-07062009-095040 Title Statistical Methods and Models for Analyzing Sugarcane (Saccharum species hybrids) Plant Breeding Data Degree Doctor of Philosophy (Ph.D.) Department Agronomy & Environmental Management Advisory Committee
Advisor Name Title Collins A. Kimbeng Committee Chair Gerald O. Myers Committee Member Kenneth A. Gravois Committee Member Kevin S. McCarter Committee Member Roberto N. Barbosa Dean's Representative Keywords
- log linear models
- cross resistance
- sugarcane borer
- mexican rice borer
- multivariate repeated measures
- seedling selection
- logistic regression models
- neural network models
- family evaluation
- random coefficient models
Date of Defense 2009-06-10 Availability unrestricted AbstractEarly generation selection of sugarcane families using means is inadequate while visual seedling selection is subjective and inefficient. Data from advanced variety trials (yield, quality and agronomic traits) are collected over several crop-years to determine yield potential and ratooning ability of genotypes follow a multivariate repeated measures structure. In Louisiana, the sugarcane borer and recently the Mexican rice borer are major insect pests of sugarcane. Both borers have similar feeding habits, providing an opportunity for investigating if genotypes resistant to one species would provide resistance to the other (cross-resistance). The objectives of the study were to identify statistical methods to evaluate family yield potential and repeatability, enhance seedling selection for yield, analyze advanced variety trials data and prove cross resistance between the sugarcane borer and the Mexican rice borer.
Random coefficient models (RCM) identified elite families with higher cane yield potential and higher repeatability between seedlings and clones. These elite families comprised a larger proportion of higher yield seedlings that produced high yielding clones. Logistic regression models (LRM) provided an objective statistical decision support tool for selecting high yielding seedlings and were more flexible at adjusting the number of seedlings to advance than visual selection. The LRM can be used to identify important traits in breeding populations as well as directionally shifting population trait values during selection. Neural network models can be used to automate the LRM. The multivariate repeated measures analysis (MRM) reduced Type I errors associated with univariate analysis by including covariance to compute experimental errors. The MRM showed greater statistical differences among genotypes for yield traits than univariate analysis. Cross resistance between the sugarcane and Mexican rice borer was proved using log linear models, and using a population with known sugarcane borer resistance status.
Using RCM will significantly increase the efficiency of early generation selection by identifying families with high yield potential and repeatability while LRM will increase efficiency of identifying high yielding seedlings from these elite families. MRM will increase the accuracy of evaluating genotypes for yield and ratooning ability. Cross-resistance will allow breeders to take advantage of parents from the sugarcane borer recurrent selection program.
Filename Size Approximate Download Time (Hours:Minutes:Seconds)
28.8 Modem 56K Modem ISDN (64 Kb) ISDN (128 Kb) Higher-speed Access Zhou_diss.pdf 2.02 Mb 00:09:19 00:04:47 00:04:11 00:02:05 00:00:10
If you have questions or technical problems, please Contact LSU-ETD Support.