NOTE: BIS 525 students are required to attend in person. Others are invited to attend in person, but may also attend via Zoom.
SPEAKER: Jinyuan Liu, PhD, Assistant Professor, Department of Biostatistics, Vanderbilt University Medical Center
TITLE: “Feature Aggregation in Causal Discovery for High-dimensional Data: Application to Targeting the “Gut-Brain-Axis” via the Microbiome Diversity"
ABSTRACT: The high-dimensional data is emerging rapidly in today’s research, incentivized by technological advancements such as high-throughput sequencing. To derive scientific insights, the feature extraction framework, such as lasso, has been ubiquitously adopted. However, appropriate inference for those “cherry-picked” features is not only difficult but has been overlooked historically. Moreover, the assumption of sparse features may not be true in practice, prompting some research areas (e.g., genetics, neuroscience) to advocate an overall composite effect, rather than locating individual culprits. Nevertheless, with the scientific frontier shifting towards causal interpretations, evaluating such associations is no longer a sufficient goal. Adopting this feature aggregation framework, we propose a causal discovery process in line with the paradigm shift to capture both within- and between-subject attributes of the entire vector of high-dimensional features. Therefore, this timely solution can provide new perspectives to target the overall causal effects via scientific-relevant metrics. Also, by pinpointing their efficient influence function (EIF), we develop the augmented doubly robust (DR) causal estimators that achieve optimal asymptotic properties without the sparsity assumption. More importantly, we offer a joint inference strategy to achieve seamless statistical inference from several processes to reduce the computational overhead substantially. This causal discovery approach was illustrated to demystify the “gut-brain-axis” through the microbiome diversity but can easily be adapted to various disciplines with burgeoning high-dimensional data.
YSPH values inclusion and access for all participants. If you have questions about accessibility or would like to request an accommodation, please contact Charmila Fernandes at firstname.lastname@example.org. We will try to provide accommodations requested by September 14, 2023.
Vanderbilt University Medical CenterJinyuan Liu, PhDAssistant Professor