Volume no :10, Issue no: 1, September (2013)

FINITE-DIFFERENCE DERIVATIVES OF A FIRST-ORDER INTEGRAL APPROXIMATION QUANTIZED WITH A DEFAULT DUAL QUASI-NEWTON OPTIMIZER AND A PSEUDO-LIPSCHITZIAN PROPERTY FOR PREDICTIVE MAPPING SPATIALLY INHOMOGENEOUS EXPLANATORY COVARIATES

Author's: BENJAMIN G. JACOB, ROBERT J. NOVAK, LAURENT TOE, MOUSSAS S. SANFO, SEMIHA CALISKAN, ROSE TINGUERIA, ALAIN PARE, LAURENT YAMEOGO, DANIEL GRIFFITH and THOMAS R. UNNASCH
Pages: [1] - [136]
Received Date: July 1, 2013
Submitted by:

Abstract

Currently the most common hypothesis-testing methods for predictive seasonal risk modeling arthropod-related infectious disease data is employing traditional multivariate analogues. Unfortunately, statistical programs are not stringent enough for seasonally quantitating error and error assumptions in the probability plots of regressed forecasts for accurately interpolating endemic transmission regions in an epidemiological study site. In this paper, we propose a Gaussian process in a spatial filter analyses and a Bayesian matrix for deriving qualitative probabilistic inferences from an ecological regressed dataset of noisy georeferenced riverine-based larval habitats of Similium damnosum, a black fly vector of onchocerciasis. Our intention was to simulate optimally unbiased seasonal endemic transmission-oriented explanatory covariate coefficients based on spatial aggregations of productive habitats within a riverine epidemiological study site by introducing a latent variable within a non-linear autoregressive equation. Autocorrelation scatterplots revealed that the Moran’s coefficient was 0.067, while the Gearys ratio was 0.891 in the forecasts. Improvement of fit of a WinBUGS hierarchical Bayesian model then revealed that the adjusted covariate Presence of hanging vegetation was statistically important to prolific sampled habitats.

Keywords

PROC NL MIXED, Moran’s coefficient, SAS Macro Win BUGSio, Bayesian, onchocerciasis.