Consequences of spatial heterogeneity in the analysis and design of experiments

Abstract

Consequences of spatial heterogeneity in the analysis and design of experiments. Many traits assessed in field ecological trials show  nonrandom  spatial  structures  that  may  affect  the  efficiency  of  standard  statistical  analyses.  Although  several  more  or  less
sophisticated experimental designs may improve this efficiency by controlling the spatial variation, there are many situations where designs can not be properly arranged to actual spatial patterns. In such cases, spatial analysis techniques become essential to correctly analyse spatial  autocorrelated  data.  In  this  paper,  the  effects  of  spatial  autocorrelation  on  the  results  of  conventional  statistical  analysis  are discussed, and a spatial adjustment procedure, based on geostatistics, is proposed to be used when data are spatially autocorrelated. A case study is presented to show how conventional analysis of spatially autocorrelated data may give completely erroneous conclusions.