Justin Garrish's recent article published

Dr. Justin Garrish, in his second year as a postdoc at the department, has published a new article:

Garrish J., Nychka D., Chan C.L., and Diniz Behn, C. Incorporating Physiology to Improve Estimates of Post-prandial Insulin Secretion Rates with Quantified Uncertainty. SIAM Journal on Applied Mathematics. 2026; 86(4): 1422-1440.

Here is a brief non-technical summary:

We develop a Bayesian hierarchical model for estimating insulin secretion rate (ISR), the rate at which the pancreas releases insulin. We model ISR on a logarithmic scale, which guarantees that the estimated secretion rate is positive and also provides a natural way to describe its uncertainty. We use a flexible quadratic trend to represent the typical rise and fall of insulin secretion during a glucose tolerance test.

Our model improves on an earlier Bayesian model that could produce physiologically impossible negative estimates of insulin secretion. We use a Newton–Raphson numerical method to efficiently estimate the model parameters and quantify their uncertainty. We apply the new model to oral glucose tolerance test data from young people with and without cystic fibrosis. 

By incorporating the known physiological requirement that insulin secretion must be positive directly into the statistical model, our approach provides a more reliable and precise way to estimate insulin secretion and may improve the assessment of pancreatic beta-cell function in both health and disease.