Re: Time as a dummy variable

[Dave Reilly]

A.

 

A deterministic component such as "trend" or period to period growth an arise in two distinct ways

 

1. y(t)=y(t-1)+ trend   where for example the coefficient trend may vary over time e.g.

 

         t= 1,....t1       trend =tr1

         t=t1+1,.....t2  trend=t2

 

    this model can include auxiliary regressors x1,x2,.... xP  ( and any appropriate lags of the P helping series ) and is referred

    to as a Transfer Function Model with a steady state differential or slope whose value may need to be localized .

 

 

a second way

 

 

2. y(t)= w1*X where X =1,2,.......t

 

    this model can include auxiliary regressors x1,x2,.... xP  ( and any appropriate lags of the P helping series ) and is referred

    to as a Transfer Function Model with a deterministic input series X .

 

    Now it is very clear that there may be NO TREND for the first t1 values and one trend thereafter ( for example ) thus one would form

 

    X=0 for i=1,....t1-1

        1 for i=t1

        2     i=t1+1

        T-t1 i=T

 

    thus estimating the growth over the valid range t=t1 to t=T

 

    Of course there could be multiple break points in TREND

 

 

Summary:

 

 

Early work in econometrics assumed either MODEL 1 OR 2 and furthermore assumed one trend coefficient .

 

Work in the area of Intervention Detection started about 1979 under George Box and George Tiao at Wisconsin and has

been fully implemented into our software AUTOBOX and our freeware offering FreeFore .

 

 

http://www.autobox.com

 

 

Hope this helps .... As it should precede the intent of your question ...

 

 

dave reilly

Automatic Forecasting Systems

215-675-0652

 

 

 

 

  

[Dave Reilly] 

 

 -----Original Message-----
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]On Behalf Of Anupama Ramachandran
Sent: Thursday, March 25, 2004 1:22 AM
To: [EMAIL PROTECTED]
Subject: [edstat] Time as a dummy variable

Hi

I have a utility model in which time is treated as a dummy variable with values 0 and 1. In the utility model, time will interact with at least two other terms. If I specify time to be a dummy variable are interactions allowed with dummy variables? What are the specifics I should follow while doing this? Are their any papers which dela with the treatment of time as a dummy variable?

I'd really appreciate any help or suggestions in the same.

 

Thanks

Anupama

Anupama Ramachandran

Ph.D Student

Dept. of marketing

R.H. Smith School of Business

University of Maryland, College Park