[EMAIL PROTECTED] (Norm Loomer) wrote in message news:<[EMAIL PROTECTED]>... > A colleague brought me some annual time series data that shows a > downward trend from 1992 to 2001. In 1996 the agency that recorded the > data adopted a new program that, if successful, would cause the data to > trend downward more sharply. What they are looking for, I think, is > evidence that the slope after 1996 is greater (in absolute value) than > the slope before 1996. Can someone help me with this? > > -Norm Loomer > > . > . > ================================================================= > Instructions for joining and leaving this list, remarks about the > problem of INAPPROPRIATE MESSAGES, and archives are available at: > . http://jse.stat.ncsu.edu/ . > =================================================================
Norm, A comprehensive approach to a time series model would investigate the need for three kinds of structures ( http://www.autobox.com/blp004.html ) 1. AUTO-PROJECTIVE component reflecting the impact of omitted stochastic causal series 2. Identified INTERVENTIONS or dummy variables reflecting omitted deterministic causal series. 3. Causal Series , their lags and sometimes their leads. Intervention Variables are essentially 0,1 variables that can have different formats viz. a. A Pulse Variable reflecting a one time event ( 0,0,0,....,0,1,,0,0,....0) b. A Seasonal Pulse reflecting a systematic seasonal event ( 0,0,1,0,0,0,1,0,0,0,1,0,0,0,1,..... c. A STEP or LEVEL SHIFT reflecting a permanent effect that started at a point in time ( 0,0,0,0,0,1,1,1,1,1,.....1) thus reflecting a change in intercept. d. A TIME TREND reflecting a change in slope Note that B is the Backshift Operator .....AKA as L for Lag Operator to Econometricians .... (1-B)L=P thus when you difference a LEVEL SHIFT variable you get a PULSE thus P=L/(1-B) Now a LEVEL SHIFT is simply a difference of a TIME TREND Consider 0,0,0,0,1,2,3,4,5,6,7,8,......T a TIME TREND that started at t=5 if you difference this variable or series you get 0,0,0,0,1,1,1,1,1,1,1,1,......1 thus a L=(1-B)T or T=L/(1-B) or T=P/[(1-B)(1-B)] AUTOBOX ( http://www.autobox.com/abx5.exe ) and FreeFore it's ShareWare clone ( http://www.autobox.com/freef.exe ) have a fully implemented search process to detect and incorporate all of these Intervention Variables in it's tour de force. When you download FreeFore carefully examine two time series examples which illustrate this powerful feature ( scrptbj1.asc and scrptbj2.asc ....monthly US prescriptions relevant to a major disease ) Of course if you have a lot of time on your hands you could program the search process using the software tools that you currently hace access to. Intervention Detection is laid out in a few URLS'S off of http://www.autobox.com/teach.html I would be glad to take your time series and use FreeFore to analyze it or to help you use FreeFore . Hope this helps.. Remember what orson Bean said about trends.... " A trend is a trend until it bends ....etc ... " Regards Dave Reilly Senior Vice-President Automatic Forecasting Systems http://www.autobox.com 215-675-0652 . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
