##### Department of Mathematics,

University of California San Diego

****************************

### Statistics Seminar

## Richard Song

#### UC Berkeley

## High Dimensional Time Series Modeling

##### Abstract:

\indent We usually only justify time series estimators using asymptotic theory, but the sample size for time series, say those yearly macro series, is usually limited, not more than 100. Additionally, high dimensionality and serial dependence makes the asymptotics harder to be a good approximation for a finite sample. My works in high dimensional time series modeling tries to solve these problems, i.e. to quantify the interplay and strike a balance between degree of time dependence, high dimensionality and moderate sample size (relative to dimensionality). In this talk, I will talk about generalized dynamic factor models (briefly), large vector autoregressions for modeling expectation (in detail), and also dynamic volatility matrix estimation (briefly) if time permits.

### February 14, 2011

### 1:00 PM

### AP&M 6402

****************************