Abstract

MORALES, Victor; MORA, Argenis; GARAY, Vicente  and  VALERA, Lino. Spatial variance analysis as an alternative for assessment of forest progeny tests. Agron. Costarricense [online]. 2016, vol.40, n.2, pp.87-105. ISSN 0377-9424.  http://dx.doi.org/10.15517/rac.v40i2.27388.

The use of block design and randomization of different progenies in each block may be insufficient for local variation control within trial area. The presence of dispersed micro-sites influences the growth of individuals, significantly impacting conventional variance analysis results. The present study proposes and describes a methodology of analysis for progeny test in forest research, comparing various spatial variance analysis methods with reference to incorporating or not incorporating the block effect and the modeling of surface trend on data recorded from a trial of Pachira quinata (Jacq.) WS Alverson progenies over a period of 3 years. A protocol analysis was established which includes diagnosis of residual correlation, modeling of the structure of correlation-variation and a spatial variance analysis. The existence of residual or experimental errors, that do not meet the fundamental requirement of independence assumption in order to do an analysis of variance based on design, was demonstrated. The best-fit model was the one that captured the surface variation using the cubic polynomial with the modeling of correlated errors (spherical geostatistical model). Progeny test means estimated by the spatial model with polynomial trend showed a different order of the test means estimated under the analysis based on block design.

Keywords : Progeny trials; spatial analysis of variance; geostatistical analysis; mixed linear models; Pachira quinata.

        · abstract in Spanish     · text in Spanish     · Spanish ( pdf )