DATE:

Friday, October 20, 2023

TIME:

3:30 PM

LOCATION:

GMCS 314

SPEAKER:

Xiyue Liao, Mathematics and Statistics, San Diego State University

ABSTRACT:

Shape- and order-constrained estimation is an important nonparametric method which only has vague shape or order assumptions such as monotonicity or convexity. Because it minimizes assumptions about the underlying trend or pattern in a data set, it can effectively guard against model mis-specification errors. In this talk, we will give a brief overview of how this framework works and illustrate it with three applications: 1) change-point detection in the time series of Landsat imagery monitoring annual forest disturbance dynamics 2) estimation of the curve of Luteinizing Hormone (LH) concentration in picograms of ewes. 3) isotonic domain mean estimation of the cholesterol level with the National Health and Nutrition Examination Survey data.

Bio: Xiyue Liao is a statistician, with both applied and theoretical interests, which include shape- and order-constrained estimation and inference, statistical software development, machine learning, etc. Dr. Liao received her PhD degree in Statistics in 2016 from the Department of Statistics at Colorado State University. Before joining SDSU, she was an Assistant Professor in the Department of Mathematics and Statistics at CSULB, mainly responsible for designing and teaching upper-division data science courses and advising graduate students’ theses and projects. Before that, she was a Postdoc Fellow in the Department of Statistics and Applied Probability at UCSB, and she was in charge of project-based mentoring courses that apply machine learning methods to build predictive models in R or Python with health insurance data sets provided by industry partners.

HOST:

Parag Katira

VIDEO: