| Abstract: | Negative spatial autocorrelation (NSA), the tendency for dissimilar neighboring values to cluster on a map, may go undetected in statistical analyses of immature Anopheles gambiae s.l., a leading malaria mosquito vector in Sub-Saharan Africa. Unquantified NSA generated from an inverse variance-covariance matrix may generate misspecifications in an An. gambiae s.l. habitat model. In this research, we used an eigenfunction decomposition algorithm based on a modified geographic connectivity matrix to compute the Moran's I statistic, to uncover hidden NSA in a dataset of georeferenced An. gambiae s.l. habitat explanatory predictor variables spatiotemporally sampled in Malindi and Kisumu, Kenya. The Moran's I statistic was decomposed into orthogonal synthetic map patterns. Global tests revealed that |zMC|s generated were less than 1.11 for the presence of latent autocorrelation. The algorithm captured NSA in the An. gambiae s.l. habitat data by quantifying all non-normal random variables, space-time heterogeneity, and distributional properties of the spatial filters. |
