English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Computer-coded verbal autopsy (CCVA) algorithms predict cause of death fromhigh-dimensional family questionnaire data (verbal autopsies) of a deceasedindividual. CCVA algorithms are typically trained on non-local data, then usedto generate national and regional estimates of cause-specific mortalityfractions. These estimates may be inaccurate if the non-local training data isdifferent from the local population of interest. This problem is a special caseof transfer learning. However, most transfer learning classification approachesare concerned with individual (e.g. a person's) classification within a targetdomain (e.g. a particular population) with training performed in data from asource domain. Epidemiologists are often more interested in estimatingpopulation-level etiological distributions, using datasets much smaller thanthose used in common transfer learning applications. We present a parsimonioushierarchical Bayesian transfer learning framework to directly estimatepopulation-level class probabilities in a target domain. To address smallsample sizes, we introduce a novel shrinkage prior for the transfer error ratesguaranteeing that, in absence of any labeled target domain data or when thebaseline classifier has zero transfer error, the calibrated estimate of classprobabilities coincides with the naive estimates from the baseline classifier,thereby subsuming the default practice as a special case. A novel Gibbs samplerusing data-augmentation enables fast implementation. We extend our approach touse not one, but an ensemble of baseline classifiers. Theoretical and empiricalresults demonstrate how the ensemble model favors the most accurate baselineclassifier. We present extensions allowing class probabilities to vary withcovariates, and an EM-algorithm-based MAP estimation. An R-package implementingthis method is developed.

Regularized Bayesian transfer learning for population level etiological distributions