Count data, ordered data, binary data
My interest in categorical data dates back to my first year in grad school where I became fascinated with matched proportions in particular. Count data, ordered data, binary data, all present interesting challenges to the modeling process.
Sparse Multivariate Matched Proportions
I developed the first Bayesian multivariate matched proportions model that is also the first to examine and address the impact of sparse response. This work leverages a functional data tool to model a latent covariance matrix.
Ordinal Function-on-Scalar Regression
This work extends function-on-scalar regression models from two different frameworks, wavelet-based and penalized spline-based, to the ordinal setting. To fit these models we leverage the probit representation of the ordinal model, performing the basis transformations in the latent-space.
Non-informative priors for COM-Poisson
This work investigates non- and weakly-informative prior for the COM-Poisson distribution. We derive the Jeffreys’ prior and develop several other weakly- to non-informative priors based off of the conjugate prior. A pre-print is available on arXiv.
Other Categorical Work
I’ve also done work on two-phase tests for genome wide associations scans, sensitivity and specificity adjusted prevalence estimation, ordinal lagged effects, and permutation tests for Goodman-Kruskal’s Gamma.