PhD Student, University of California - Riverside
Partial Overlapping Batch Means Estimators
Batch Means estimators are useful for estimating the sample mean variance when dealing with correlated observations. The most accurate method is overlapping batch means, but it becomes computationally complex as the number of observations increases. The non-overlapping batch means estimator reduces computation time but increases variability. We propose weighted partial overlapping batch means with different amounts of overlapping giving practitioners the flexibility to determine the tradeoff between variability and computational complexity. These new estimators can be applied in time series, econometrics, steady-state simulation, Markov chain Monte Carlo, and stochastic gradient descent.
My name is Noe Vidales. I am a fourth year PhD Candidate at UCR. My research interest lies in stochastic processes. Specifically, I am interested in Markov Chains and convergence complexity. Outside of academia, I love cycling and finding scenic oceanside trails in Southern California. I also love learning about and discovering new cultures through their cuisine.