Analysis of Texas LoanSTAR Data

Abstract

The analysis effort during the first year of the Texas LoanSTAR Monitoring and Analysis Program has emphasized selection and development of baseline analysis techniques to cover the range of buildings expected in the program. PRISM has been adopted as the baseline technique for buildings which are appropriate for treatment with one-, three- and five- parameter segmented linear, change-point models. In addition to PRISM, two- and four- parameter linear, segmented change-point models are expected to be suitable for at least preliminary analysis of monthly and daily data for all buildings in the program. Regression analysis with hourly scheduling profiles will be used for baseline analysis of hourly data. Substantial effort has been devoted to exploratory analysis intended to refine the analysis performed with the baseline techniques. Work to date has centered on investigation of Principal Component Analysis, an improved goodness-of-fit indicator for n-parameter change-point models, and calibrated simulation modeling. Data from five buildings is used to explain and illustrate the baseline analysis techniques and the exploratory work conducted.