Daniel Schmidt is a Senior Lecturer in Data Science at the Faculty of Information Technology, Monash University, Melbourne, Australia.

Since completing his PhD in information theoretic inference of linear time series models he has spent 10 years working primarily in the area of Bayesian inference, with specific interest in applications to epidemiological problems, particular the area of statistical genomics.

His specific research interests include:

• Bayesian inference of high dimensional regression models, particularly linear and generalized linear models. He has, along with Dr. Enes Makalic, written a highly efficient toolbox supporting state-of-the-art Bayesian shrinkage priors for high dimensional regression models (available here);
• Information theoretic statistics, particularly the application of information theory to statistical inference through the Minimum Message/Description Length principles;
• Statistical genomics, risk prediction and variant discovery, particularly in the areas of cancer genomics.

He is interested in using mammography and machine learning techniques to improve risk prediction and the stratification of women by their future risk of breast cancer, with the aim of assisting the creation of personalised screening programmes.

 
Recent pre-prints:

•  Adaptive Bayesian Shrinkage Estimation Using Log-Scale Shrinkage Priors, which describes a new class of shrinkage prior distributions for regression coefficients that can adapt to the sparsity characteristics of the underlying regression coefficients. It also provides simple bounds on behaviour of most existing Bayesian shrinkage priors.
• A Minimum Message Length Criterion for Robust Linear Regression, which develops a simple, finite sample criterion for model selection in linear models with heavy tailed error distributions using the Minimum Message Length principle. Interestingly, the penalty term can be shown to be related to the signal-to-noise ratio of the fitted model.

External Links:

• His personal github page is https://github.com/dfschmidt80.
• The BayesReg package for efficient, high dimensional Bayesian penalized regression can be downloaded from here.
• His google scholar page is here.

Daniel Schmidt has developed, and acted as Chief Examiner and Lecturer, for the following units at the Faculty of Information Technology:

• FIT2086 Modelling for Data Analysis
• FIT3154 Advanced Data Analysis