Hierarchical Bayesian Methods for Alzheimer`s Disease

Discover how to access and upload longitudinal Alzheimer's disease datasets from the OASIS and Kaggle platforms, preparing for bayesian multilevel analyses.

Explore knowledge discovery through data exploration of brain deficits, focusing on age and normalized whole brain volume, using box plots, scatter plots, and a simple single-level model.

Explore sources of variation in brain volume data using hierarchical Bayesian methods, grouping by dementia status, CVR rating, gender, education, and socio-economic status to model age-related brain atrophy.

This lecture demonstrates data categorizations for features in Alzheimer’s disease research, using finite intervals for education and socioeconomic status and binary coding for gender and handedness.

Learn how to identify and impute missing values using the missForest algorithm in a multilevel Bayesian analysis, including categorizing non-numeric group features and reassembling the complete dataset.

Apply a fixed-effect Bayesian single-level analysis in Stan to Alzheimer’s data, using age as predictor and normalized brain volume as the response, formatted as a Stan data list.

Construct a Stan fixed-effect model: a linear relation of normalized whole brain volume on age, with intercept, slope, and sigma, plus data, parameters, model, and generated quantities blocks.

Execute a fixed effect Bayesian model in Stan, input data, run iterations, verify convergence with trace plots and R-hat, and extract intercept and slope for brain atrophy after age 60.

Visualize the posterior uncertainty of deficits by plotting deficits against age, drawing many posterior lines to reveal the spread and uncertainty with color-coded intercepts.

Perform a posterior predictive check for a fixed effect Bayesian model, comparing observed and replicated brain volume data to assess model fit and age-related decline in Alzheimer's disease.

Explore hierarchical Bayesian analysis of brain atrophy across age 60–90, comparing converted, demented, and non-demented groups, using a data list and blocks for Stan model B.

Explore constructing random intercepts with group-level variation in a hierarchical Bayesian framework for Alzheimer's disease, defining group-specific intercepts, priors, and transformed parameters to capture dementia status effects.

Create a hierarchical Stan model for Alzheimer's disease by defining priors, specifying a normal likelihood for normalized whole brain volume, and estimating group-level random intercept variation.

Execute the hierarchical stan model to estimate random intercepts for converted, demented, and non demented groups, confirm convergence with trace plots and arhat statistics, and draw from the posterior.

Explore hierarchical Bayesian methods to model brain atrophy at age 60 with random intercepts across converted, demented, and non-demented groups, and interpret posterior estimates and credible intervals.

Explore how random intercepts and random slopes reveal group-level brain-volume changes with age in Alzheimer's disease, comparing demented and non-demented groups.

Explore a random slope in a hierarchical bayesian model to capture brain atrophy differences between demented and non demented groups using Stan blocks.

Extend a hierarchical Bayesian model by adding a level for Alzheimer’s stages, using four cognitive test levels to reveal stage-dependent brain atrophy slopes in demented patients, implemented in R.

Add a new hierarchical level to the data to model brain-size changes across Alzheimer's stages, converting stage scores into factors and enabling four levels.

Extend Stan blocks by adding a new hierarchical level for CVR levels. Update data and parameters blocks, implement a W matrix for random intercepts and slopes, and specify priors.

Explore hierarchical Bayesian methods to estimate brain atrophy rates across demented and non demented groups in Alzheimer's disease, using Stan, convergence checks, and posterior models of random intercepts and slopes.

Extend Bayesian hierarchical models for Alzheimer's disease by adding gender, education, and socio-economic level, and learn to structure data and code to sample from the posterior distribution despite computational demands.

Master the complete Stan model for hierarchical bayesian methods in Alzheimer's disease, detailing data blocks, random intercepts and slopes, and priors for multilevel analysis.