Ph.D. Candidate: Sean McLaughlin
Research Advisor: Risa Wechsler
Zoom Link: https://stanford.zoom.us/j/91777804167
Email email@example.com for password
Title: Improving Inference of Cosmological Parameters with Advanced Statistical Techniques Using Simulations
The big questions posed in cosmology, like "What is the Universe made of?" and "How did it evolve to its present state?", are fundamentally statistical in nature. As cosmological surveys continue to amass larger quantities of data and make increasingly more precise measurements, our analyses must also become more advanced, to make optimal use of the information available. In this thesis defense, I will present several efforts focused on improving the precision, accuracy, and interpretation of cosmological inferences. I will first describe my work with cosmic emulators, models that predict how galaxy summary statistics change as a function of cosmology as well as galaxy formation parameters. Using these emulators, I will show how the existence of galaxy assembly bias affects cosmological inference, and how the introduction of new models and summary statistics can improve inferences in these contexts. Finally, I will show the results of my work to isolate the features used by cosmological neural networks to constrain cosmology and provide evidence that these advanced machine learning models mimic techniques already well understood by cosmologists.