Takenori SAKUMURA, Ph.D.

Lecturer, Department of Industrial and Systems Engineering, Hosei University / 日本語

Research Topics

Bayesian estimation with noninformative priors

We derive Bayesian estimators that are optimal from the viewpoint of prediction under noninformative priors (reference and Jeffreys priors), and study their properties for small samples. Distributions include the von Mises, Weibull, gamma, extreme value and Birnbaum–Saunders distributions.

Keywords: reference prior, Jeffreys prior, canonical parameter, Kullback–Leibler loss, approximate unbiasedness

Selected papers:

Reliability, degradation and accelerated life tests

We predict product lifetime from tests under raised stresses such as temperature and voltage, and from degradation data. Topics include cumulative exposure models, Weibull models with multiple stresses, degradation models based on Wiener and inverse Gaussian processes, and their Bayesian estimation and optimal test plans.

Keywords: accelerated life test, cumulative exposure model, dual-stress Weibull model, degradation model, optimal test plan

Selected papers:

Item response theory and adaptive testing

Based on item response theory, we study adaptive tests that estimate ability from responses and choose the next item accordingly. Topics include multistage tests, item selection criteria and test equating and linking.

Keywords: item response theory, adaptive testing, multistage test, equating and linking, Bayesian estimation

Selected papers:

Statistical modeling of real data

We express the structure behind real data with statistical models, including joint estimation of incubation periods and infection times, demand forecasting of parts considering units in operation, analysis of purchase histories and visit intervals, and sports data analysis.

Keywords: survival analysis, censored data, demand forecasting, sports data

Selected papers:

back to top of page