[EMAIL PROTECTED] (Donald Burrill) wrote in message news:<[EMAIL PROTECTED]>... > I'm not familiar with the "Chow test". But your problem is logically > equivalent to fitting the model > > Rit = constant + b1RMt + b2TWNt + > b3In(constant + b1RMt + b2TWNt) + > b4Post(constant + b1RMt + b2TWNt) > > where In = 1 if the date lies between October 1990 and September 1992, > and In = 0 otherwise; and Post = 1 if between October 1992 and December > 2001, and Post = 0 otherwise. (Of course, what you actually do is > construct In and Post, then X4 = In*RMt, X5 = In*TWNt, X7 = Post*RMt, > X8 = Post*TWNt, and fit the model > > Rit = b0 + b1RMt + b2TWNt + b3In + b4X4 + b5X5 + b6Post + B7X7 + b8X8 > > If any of the six coefficients (b3, b4, b5, b6, b7, b8) differ(s) > significantly from zero, you have evidence of a structural break in the > model. If none of them differs significantly from zero, there may or > may not be a structural break, but if there is it's too subtle to be > detected in the current data.) > > On Wed, 13 Aug 2003, Ahmed El-Masry wrote: > > > I am doing regression as follows: > > Rit = constant +b1RMt + b2TWNt > > My sample period ranges from January 1981 to December 2001 and my > > sample is 364 companies and the data is monthly. > > I would like to test whether there is a structural break in the model, > > so I partion the sample period to three specific subperiods which are: > > 1- January 1981 to September 1990 (Pre); > > 2- October 1990 to September 1992 (In) and > > 3- October 1992 to December 2001 (Post). > > Can anybody show me how I can do this test by Chow Test using SaS. > > ----------------------------------------------------------------------- > Donald F. Burrill [EMAIL PROTECTED] > 56 Sebbins Pond Drive, Bedford, NH 03110 (603) 626-0816 > > . > . > ================================================================= > Instructions for joining and leaving this list, remarks about the > problem of INAPPROPRIATE MESSAGES, and archives are available at: > . http://jse.stat.ncsu.edu/ . > =================================================================
http://www.princeton.edu/~gchow/ The Chow Test is as you surmized a test for the equivalence of coefficients ...and not a test for individual coefficients .. WE use it in AUTOBOX and FREEFORE to find the break-points in time where the coefficients differ significantly from one another .. Regards Dave R AFS . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
