To accommodate overdispersion we extended Poison model to conjugate Poisson-Gamma model. <...> In conclusion we admit the auspicious features of Bayes hierarchical modeling with powerful fl exibility to disclose underlying patterns in data distribution. <...> We proved that there is conspicuous space pattern in risk of contagion with onychomycosis. <...> The space structured residuals revealed signifi cant sigma (s) estimates that in both space specifi cations leaves zero beyond 95% CI. <...> This in tune with model fi t criterions bears witness to signifi cance of space component in aggregated counts distribution and support space nature of data. <...> Dynamic eff ect also appeared to be signifi cant with and indicates the decline in risk of contagion in 2011–2013 by factor of 0.755. <...> Our data rebut hypothesis on infl uence of population density as good proxy for development of territory on risk of onychomycosis infection. <...> Yet the study sustained the number of saunas and beauty salons as signifi cant risk factors of infection. <...> Interesting that soil pH also exerted its infl uence admittedly due to inhibition of pectinolytic activity of soil micro fl ora. <...> Number of district population eff ect was insignifi cant under any specifi cation. <...> The most powerful and fl exible recognized to be Bayes approach realized through Monte Carlo Markov Chain (MCMC) samplers [1]. <...> Given aggregated by districts annual counts of onychomycosis incidence we opted for classical discreet Poisson space model. <...> We also use number of district population to make allowance for population density. <...> Another factor was weighted soil pH value. <...> Basic Poisson overdispersed model We discovered that classical Poisson model failed to describe observed counts for the latter is over dispersed that is often the case for aggregated data [2]. <...> To accommodate overdispersion we extended Poison model to conjugate Poisson-Gamma model: ( , ) yi ~ i ~ Gamma a Poisson ( ) i i i First two moments of expected counts µі distribution parameters і =аі/і that resulting dispersion of observed counts (уі Var(уі)=E[Var(уі | µі)]+Var[E(уі | µі <...>