PRIISM “Statistics in Society” Lecture – “Forecasting Large Panel Data with Penalized Least-Squares”

by cp2530 on September 13, 2010

Title: PRIISM “Statistics in Society” Lecture – “Forecasting Large Panel Data with Penalized Least-Squares”
Location: New York University, Kaufman Management Center, Room KMC 5-80, 44 West 4th Street, New York City
Description: We would like to bring to your attention the following talk, a PRIISM Center “Statistics in Society” lecture, co-sponsored by Stern’s IOMS Department, THIS FRIDAY AT NOON:

Professor Jianqing Fan (Princeton University), will present his paper “Forecasting Large Panel Data with Penalized Least-Squares” on Friday, September 17, 12-1pm, at Kaufman Management Center, room KMC 5-80.

Large Panel data arise from many diverse fields such as economics, finance, meteorology, energy demand management and ecology where spatial-temporal data are collected. Neighborhood correlations allow us to better forecast future outcomes, yet neighborhood selection becomes an important and challenging task. In this talk, we introduce the penalized least-squares to select the neighborhood variables that have an impact on the forecasting power. An iterative two-scale approach will be introduced. The inherent error (noise level) will also be estimated in the high-dimensional regression problems, which serves as the benchmark for forecasting errors. The techniques will be illustrated in forecasting the US house price indices at various Core Based Statistical Area (CBSA) levels.
Start Time: 12:00
Date: 2010-09-17
End Time: 13:00

Previous post:

Next post: