Experimental design and data collection for inferring properties of heterogeneous population

Data availability, such as the type, quantity, and quality of data, determines and often restricts the use of mathematical models. Thus, planning the data collection ahead is important to better utilize the mathematical models in studying specific research questions. In this talk, I will present some of my work on designing experiments and data collection to infer the interaction between heterogeneous populations robustly. First, experimental designs of mono-culture and co-culture data are studied to infer interaction types using the Lokta-Volterra model. We conclude that collecting two mixtures of co-culture data is more robust than using mono-culture and one mixture of co-culture data despite requiring less data. Second, I will present an inference method to separately estimate the intra- and inter-species interaction in the birth and death process. By quantifying the higher statistical moments of the trajectories,  the regulation of birth or death can be distinguished, which cannot be done only by using the mean trajectory. If I have time, I will also present an information-theoretic approach to determine data collection plans to learn specific model parameters applied to obtaining patient-specific scanning schedules.

Bio
https://heyrim.github.io/