Independence between observations is a standard prerequisite of traditional statistical tests. This condition is violated in the case of variables sampled in space, such as those provided by remote sensing or geographical databases, when autocorrelation is present within the data. To overcome this spatial dependence, a team of researchers from the IRD and CIRAD, working on the Pl@ntNet project, recently developed a new method, consisting in comparing the results of associations observed in nature with those obtained from artificial data that replicate the spatial structures seen in observations. It is a method whose field of application goes well beyond ecological issues.
via CIRAD - Actualités / News http://www.cirad.fr/news/all-news-items/articles/2013/science/overcoming-spatial-dependence-in-analyses-of-biological-processes
No comments:
Post a Comment