ESTIMATION FOR THE MULTI-CONSEQUENCE INTERVENTION MODEL

• NCJ Number: 60685
• Author(s): F B ALT; S J DEUTSCH; J J GOODE
• Date Published: Unknown
• Length: 6 pages
• Grant Number(s): 75-NI-99-0091
• Sponsoring Agency: US Dept of Justice; US Dept of Justice NIJ Pub
• Corporate Author: Georgia Institute of Technology
• Sale Source: National Institute of Justice/
• Language: English
• Country: United States of America

Annotation

PROPERTIES OF THE MULTICONSEQUENCE INTERVENTION MODEL, APPLIED TO ASSESS CONTINUOUS INTERVENTION SITUATIONS, ARE DESCRIBED, WITH MAXIMUM LIKELIHOOD AND ITERATIVE CONDITIONAL LEASE SQUARES ESTIMATION TECHNIQUES.

Abstract

TIME-SERIES EXPERIMENTATION AND ITS STATISTICAL INFERENCE, FIRST INTRODUCED BY BOX AND TIAO, WAS EXTENDED BY GLASS, WILLSON, AND GOTTMAN. MODEL FORMULATIONS OF THE LATTER ASSUME THAT AUTOREGRESSIVE AND MOVING AVERAGE PARAMETERS BEFORE AN INTERVENTION ARE THE SAME AS PARAMETERS AFTER AN INTERVENTION, WHEN SUCH PARAMETERS DESCRIBE A CORRELATIVE STRUCTURE. IN THE ARTICLE, THESE FORMULATIONS ARE MADE EVEN MORE FLEXIBLE TO ALLOW FOR THE CONSEQUENCES OF AN INTERVENTION AFFECTING PROCESS-LEVEL AND CORRELATIVE PARAMETERS. IT IS SHOWN THAT MAXIMUM LIKELIHOOD AND INTERACTIVE CONDITIONAL LEAST-SQUARES ESTIMATION TECHNIQUES CAN DESCRIBE PROCESS-LEVEL AND INTERNAL CORRELATIVE STRUCTURE PARAMETERS. ALSO, MAXIMUM LIKELIHOOD ESTIMATES CAN BE USED IN SETTING UP AN ASYMPTOTIC LIKELIHOOD RATIO TEST TO INVESTIGATE THE HYPOTHESIS THAT AUTOREGRESSIVE AND MOVING AGERAGE PARAMETERS PRIOR TO AN INTERVENTION ARE EQUAL TO THOSE AFTER AN INTERVENTION. AN EXAMPLE OF THE MULTICONSEQUENCE INTERVENTION MODEL IS PRESENTED. SUPPORTING EQUATIONS AND REFERENCES ARE INCLUDED. (DEP)