Showing papers on "Nested sampling algorithm published in 2003"

Bayesian model selection for mixtures of structural equation models with an unknown number of components.

Abstract: This paper considers mixtures of structural equation models with an unknown number of components. A Bayesian model selection approach is developed based on the Bayes factor. A procedure for computing the Bayes factor is developed via path sampling, which has a number of nice features. The key idea is to construct a continuous path linking the competing models; then the Bayes factor can be estimated efficiently via grids in [0, 1] and simulated observations that are generated by the Gibbs sampler from the posterior distribution. Bayesian estimates of the structural parameters, latent variables, as well as other statistics can be produced as by-products. The properties and merits of the proposed procedure are discussed and illustrated by means of a simulation study and a real example.

A comparison of nested analysis of variance (ANOVA) and variograms for characterizing grassland spatial structure under a limited sampling budget

Abstract: It is often necessary to estimate fine-scale grassland heterogeneity in situ using ground-based spectral radiometers. However, the sampling techniques used to describe spatiotemporal heterogeneity will strongly influence perceived landscape structure. We hypothesize that nested sampling schemes based on random-effects analysis of variance (ANOVA) will provide a more reliable characterization of spatial structure than that provided by spatial dependence models of geostatistics, whose basic assumption of stationarity is often violated over patchy landscapes. To test this hypothesis, we simulated a variety of nonstationary landscapes with varying complexity of patchiness, then compared the consistency of findings of both approaches. Our results showed that significance tests by distance classes of nested ANOVA consistently provided a more stable characterization of structure than that provided by variogram parameters for all landscapes. Despite its limited scope, this simulation suggests that much more atten...

Simulations to optimize sampling of aeolian sediment transport in space and time for mapping

Abstract: The sampling frequency in space and time is often inadequate to estimate the accuracy and precision of aeolian sediment transport. The problem stems from a lack of knowledge about the spatial and temporal scale of variation in aeolian transport and is compounded by a shortage of resources (aeolian sediment traps and labour). This study developed a geostatistical methodology for estimating sediment transport at unsampled locations and tested the extent to which it was dependent on sampling networks (nested, grid and random) and frameworks (mobile or static sampling framework between wind erosion events). Aeolian transport data were collected in an area of Australia influenced by wind erosion (Diamantina Lakes National Park, southwestern Queensland) to evaluate the combination of events used for mapping transport. Insufficient wind erosion events occurred to test sediment sampling strategies and hence simulated sampling was conducted using maps of sediment transport produced with existing models of aeolian sediment transport in the same study area. Independent validation data were used to test the estimation performance. The results suggested that sampling networks that did not include information on the spatial scale of variation (i.e. grid and random sampling) did not represent adequately the sediment transport population. In contrast, a bespoke nested sampling network performed consistently better than the other networks. Overall the static framework with a nested network was recommended for estimation and mapping of sediment transport with few resources and was likely to be especially important for use over large areas. This approach has the advantage of requiring only a single pooled within-event variogram for sediment transport to be used to derive the model parameters for kriging or stochastic simulation for each event. Copyright © 2003 John Wiley & Sons, Ltd.