# Search Results

## You are looking at 1 - 10 of 14 items for

- Author or Editor: George C.J. Fernandez x

- Refine by Access: All x

### George C.J. Fernandez

Tire interpretation of variety trials conducted with many genotypes (G) grown in many environments (E) is usually complicated by the presence of the significant G × E interaction. The common statistical analysis using ANOVA and linear regression techniques are often inadequate to study the complex two-way data structure. The biplot, a multivariate technique provides, a graphical representation of the interaction, which allows the response of each G in each E to be displayed in a two dimensional plot. It displays not only the configuration of G and E, but it also relates the two. The importance of biplot display is illustrated by using the tall fescue variety trial data on mean quality ratings published by the National Turfgrass Evacuation Program. The biplot displays about 60% of the information in the 24 (G) × 23 (E) data matrix. Environments TX3 and GA1 responded differently from other environments. Based on the biplot display genotypes are grouped and their significance will be discussed.

### George C.J. Fernandez

Split-plot design is a very popular experimental design in analyzing factorial treatments in horticultural experiments. Two different sizes or types of experimental units are assigned to main plot and the split-plot treatments. The SAS procedure GLM with the TEST option is commonly used to analyze the split-plot data by assigning the correct error term to test the main plot factor. In SAS GLM, no option is available to compare the two main factors within a split-plot factor. The CONTRAST tests and LSMEAN comparisons are valid only for comparing split-plot factors within a main plot treatment. The main factor standard error provided by the LSMEAN option is also incorrect. The new PROC MIXED procedure available in SAS 8.08 or above can be used to correct these problems in split-plot analysis. The analysis of split-plot experiments using the PROC MIXED is presented here.

### George C.J. Fernandez

### George C.J. Fernandez

Selection criteria for identifying genotypes with high stress tolerance and high yielding potentials were compared using a moderately stressed, (Stress intensity, [1-(mean stress yield (Y_{p̄})/mean potential yield (Y_{s̄})] 0.73) and a severely stressed (Stress intensity, 0.24) mungbean yield data sets. Selection based on tolerance (T), difference between potential yield (Y_{p}) and the yield in stress environment (Y_{s}) favored genotypes with tolerance and low yield potentials. Selection based on the mean productivity (MP), [MP=(Y_{p}+Y_{s})/2] favored the genotypes with high yielding potential. The Stress Susceptibility Index (S), (S = [(Y_{p}-Y_{s})/Y_{p}]/[(Y_{p̄}-Y_{s̄})/Y_{p̄}], also favored the low yielding and stress tolerant genotypes. These selection criteria failed to identify genotypes with high yielding and stress tolerance potentials. Thus, a selection criterion, Stress Tolerance Index (STI) is proposed here which identifies genotypes with high yield and stress tolerance potentials. The STI takes into account both stress tolerance and yield potentials. The STI is estimated as: [Y_{p}/Y_{p̄}][1-(T/Yp̄)]. The higher the value of STI for a genotype in a given stressed environment, the higher was its stress tolerance and yield potential. The interrelationships between these stress tolerance criteria are discussed by a biplot display.

### George C.J. Fernandez

The effectiveness of using moving mean covariance analysis (MMCA) rather than randomized complete-block design (RCBD) in experimental error control was compared in a large-scale mungbean *[Vigna radiata* (L.) Wilczek] yield trial. The MMCA was superior to the RCBD, since it significantly reduced the experimental error and the coefficient of variation (cv). Inclusion of five neighboring plots in the moving mean computation provided better error control. However, the estimation of optimum number of neighboring plots to be used and moving mean calculations were tedious. The feasibility of using border-row measurements such as mean plant height at 50% flowering or mean seed yield/m of row as a covariate in an analysis of covariance (BRMCA) was examined in a separate mungbean yield trial in which border rows were planted with a check cultivar. Both border-row measurements were equally effective in reducing the experimental error. However, plant height measurements were simpler than measuring seed yield. Because border-row measurements could be readily used as covariate in analysis of covariance without a need for moving mean computation from the response variable, BRMCA could be advantageous for error control in row crops yield evaluation.

### Mathura Thillainathan and George C.J. Fernandez

A user-friendly SAS statistical and graphical application to classify genotypes evaluated under multiple sites is presented. First, the test sites are classified into three environments, LOW [(_{0.01} statistic computed from a separate two-way ANOVA models for LOW, MED, and HIGH yielding environments. Using the user-friendly SAS MACRO, EXPLORGE horticulturists can effectively and quickly perform genotype classification under multi-site evaluation. The steps involved in downloading the necessary MACRO-CALL file from the author's home page [http://www.ag.unr.edu/gf] and the instructions for running the SAS MACRO are presented. The features of this graphical method and the graphics produced by the EXPLORGE MACRO are demonstrated and validated by published data.