Bayesian estimation with integrated nested Laplace approximation for binary logit mixed models

Grilli, Leonardo; Metelli, Silvia; Rampichini, Carla

doi:10.1080/00949655.2014.935377

Cited by 20 publications

(23 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The INLA approach avoids the computational burden related to typical Markov Chain Monte Carlo techniques (Gilks et al, 1998) often used to fit Bayesian spatial models and allows accurate approximations to posterior marginal distributions of the model parameters (Grilli et al, 2014).…”

Section: Multilevel Spatial Modelingmentioning

confidence: 99%

Modeling spatial effects of PM2.5 on term low birth weight in Los Angeles County

Coker

Ghosh

Jerrett

et al. 2015

Environmental Research

View full text Add to dashboard Cite

Air pollution epidemiological studies suggest that elevated exposure to fine particulate matter (PM2.5) is associated with higher prevalence of term low birth weight (TLBW). Previous studies have generally assumed the exposure-response of PM2.5 on TLBW to be the same throughout a large geographical area. Health effects related to PM2.5 exposures, however, may not be uniformly distributed spatially, creating a need for studies that explicitly investigate the spatial distribution of the exposure-response relationship between individual-level exposure to PM2.5 and TLBW. Here, we examine the overall and spatially varying exposure-response relationship between PM2.5 and TLBW throughout urban Los Angeles (LA) County, California. We estimated PM2.5 from a combination of land use regression (LUR), aerosol optical depth from remote sensing, and atmospheric modeling techniques. Exposures were assigned to LA County individual pregnancies identified from electronic birth certificates between the years 1995-2006 (N=1,359,284) provided by the California Department of Public Health. We used a single pollutant multivariate logistic regression model, with multilevel spatially structured and unstructured random effects set in a Bayesian framework to estimate global and spatially varying pollutant effects on TLBW at the census tract level. Overall, increased PM2.5 level was associated with higher prevalence of TLBW county-wide. The spatial random effects model, however, demonstrated that the exposure-response for PM2.5 and TLBW was not uniform across urban LA County. Rather, the magnitude and certainty of the exposure-response estimates for PM2.5 on log odds of TLBW were greatest in the urban core of Central and Southern LA County census tracts. These results suggest that the effects may be spatially patterned, and that simply estimating global pollutant effects obscures disparities suggested by spatial patterns of effects. Studies that incorporate spatial multilevel modeling with random coefficients allow us to identify areas where air pollutant effects on adverse birth outcomes may be most severe and policies to further reduce air pollution might be most effective.

show abstract

Section: Multilevel Spatial Modelingmentioning

confidence: 99%

Modeling spatial effects of PM2.5 on term low birth weight in Los Angeles County

Coker

Ghosh

Jerrett

et al. 2015

Environmental Research

View full text Add to dashboard Cite

show abstract

“…As usual in Bayesian software, the inla function allows us to specify the prior distribution of the precision, instead of the variance. We avoid the default gamma prior Ga(1, 0.0005) because it has a poor performance in logistic models with nested random effects [14]. For the simulation study we choose two alternative priors for the precisions: Ga(0.001,0.001), namely the standard choice in the popular BUGS software, and Ga(0.5, 0.003737) specified according to the criterion proposed by Fong et al [10], which consists in setting the parameters of the Gamma in order to obtain a given marginal distribution for the random effects.…”

Section: Estimation Methods and Prior Distributionsmentioning

confidence: 99%

Fitting logistic multilevel models with crossed random effects via Bayesian Integrated Nested Laplace Approximations: a simulation study

Grilli

Innocenti

2017

Journal of Statistical Computation and Simulation

View full text Add to dashboard Cite

Fitting cross-classified multilevel models with binary response is challenging. In this setting a promising method is Bayesian inference through Integrated Nested Laplace Approximations (INLA), which performs well in several latent variable models. Therefore we devise a systematic simulation study to assess the performance of INLA with cross-classified logistic data under different scenarios defined by the magnitude of the random effects variances, the number of observations, the number of clusters, and the degree of cross-classification. In the simulations INLA is systematically compared with the popular method of Maximum Likelihood via Laplace Approximation. By an application to the classical salamander mating data, we compare INLA with the best performing methods. Given the computational speed and the generally good performance, INLA turns out to be a valuable method for fitting the considered cross-classified models.

show abstract

“…The major disadvantage or limitation of this method is that the computational cost increases in the form of exponential function when hyper-parameters are increased [20] [21].…”

Section: Disadvantagesmentioning

confidence: 99%

A comprehensive review on recent advances in Variational Bayesian inference

Sain

Mittal

Gupta

2015

2015 International Conference on Advances in Computer Engineering and Applications

View full text Add to dashboard Cite

Variational Bayesian (VB) inference is the latest method for prediction of data or information in various processes. It provides a faster response with a reasonable accuracy as compared to the other methods (like Monte Carlo Markov Chain (MCMC) method). There is a large literature and work on prediction of data which deals with large amount of data. When data is missing, other methods, like MCMC, cannot be used as they require complete data for processing, while VB method provides the solution with missing data also with a very fast speed. Accuracy is the main limitation with VB method. Some algorithms are developed to overcome this limitation with some computational cost. SNVA, LSVB, SSVB and some others are the latest developed method which can be used to improve the accuracy.

show abstract

Bayesian estimation with integrated nested Laplace approximation for binary logit mixed models

Cited by 20 publications

References 16 publications

Modeling spatial effects of PM2.5 on term low birth weight in Los Angeles County

Modeling spatial effects of PM2.5 on term low birth weight in Los Angeles County

Fitting logistic multilevel models with crossed random effects via Bayesian Integrated Nested Laplace Approximations: a simulation study

A comprehensive review on recent advances in Variational Bayesian inference

Contact Info

Product

Resources

About