Internal and External Validation of Spatial Microsimulation Models: Small Area Estimates of Adult Obesity

摘要: Spatial microsimulation models can be used to estimate previously unknown data at the micro-level, although validation of these challenging. This paper seeks describe an approach models. Obesity in adults were estimated small area level using a static, deterministic, spatial model called SimObesity. utilised both Census 2001 and Health Survey for England 2004–2006. Regression analysis was identify covariates that strongest predictors obesity as input variables. The calibrated regression equal variance t-tests. Two methods external undertaken; aggregating coarser geographical which available, secondly cancer tumour sites known correlated obesity. output mapped statistically significant hot (cold) spots high (low) prevalence identified. Both internal showed low errors, suggesting this satisfactory simulation. Statistically cold (simulated) existed, even after adjusting age. emphasises three steps models: 1. Accurate simulations require strong correlations between variables; 2. It is essential internally validate models; 3. Use all means possible externally model.