OVERCOMING THE PROBLEM OF MISSING VALUES IN BREEDING TRIALS

• Main Authors: D. Šimić, A. Jambrović, I. Brkić
• Format: Article
• Language: English
• Published: Faculty of Agrobitechnical Sciences Osijek 2001-12-01
• Series: Poljoprivreda
• Subjects: missing values; methods; maize; hybrids; grain yield
• Online Access: http://poljoprivreda.pfos.hr/upload/publications/1_SIMIC.pdf

Missing data are very frequent in breeding experiments. It is possible to fill it systematically, being particularly important in case of a missing entry. Objective of this study was to present biometrical options of overcoming the problem of missing values for grain yield in 22 maize trials (FAO 200) planted in 2001 in Osijek, Croatia. Although this year was favourable for maize production, total of 63.6% of trials had missing values. Method of least squares was used to estimate missing data. A simulated increasing of missing values (more than 5) affect the systematic increase of genotype mean squares. This method is useful if trials consist less than 10% missing values, since the bias due, to overestimated genotype mean squares becomes intolerable.