Stat Colloquium [Virtual]: Dr. Tucker McElroy
U.S. Census Bureau
Location
Online
Date & Time
September 15, 2023, 11:00 am – 12:00 pm
Description
Title: A Framework for Time Series Aggregation and Seasonality Using Marked Point Processes
Abstract:
This project is concerned with
modeling economic flow time series with possibly meager sampling rate, i.e.,
with very few economic transactions per epoch (or time
period). We study a quadratic Cox process, defined as a conditional
Poisson random measure with intensity given by a squared Gaussian
process. Basic properties, such as existence and void probabilities, are
developed. We examine flows of the quadratic Cox process, defined as
the integral of a corresponding Marked point process over a time epoch, and
derive the mean and covariance of such flows, demonstrating that common
forms of temporal association in economic flow time series can be obtained
through the choice of covariance function in the Gaussian intensity. We
discuss applications such as seasonal adjustment and offer numerical
illustrations of the meager property of such quadratic Cox processes.
Tags: