Prev: Can't view results in html once check "View results when they are generated"
Next: Chow test with categorical vars
From: Rick Francis on 25 Feb 2010 11:10
Hi guys, If I were executing a Chow test using a model which contains
categorical vars (each with several levels), what are the appropriate df for
the numerator and denominator? For example, let's say the model is y = int +
x + z + Cat1 + Cat2, where Cat1 and Cat2 are categorical vars, each with 5
levels. Say sample 1 has 1200 obs and sample 2 has 1300 obs. Numerator df is
3 if we ignore the 5 parameters for Cat1 and the 5 parameters for Cat2.
Alternatively, df is 13 if we consider the parms for the categorical vars.
In similar fashion, the denominator df is 2494 (1200+1300-(2*3)) if we
ignore the parms for the categorical vars, and df is 2474 (1200+1300-(2*13))
if we consider the parms for the categorical vars. The sample is large
enough that the results are not significantly different either way. But want
avoid an unnecessary reviewer comment, and preserve my credibility with the
reviewer. Thank you very much for any insight! Rick
 | 
Pages: 1
Prev: Can't view results in html once check "View results when they are generated"
Next: Chow test with categorical vars