Dear all,
I post below the reponses I had from my earlier posting titled as above.
Cathy Smith.
=====================================================
From Rinus voeten
Don't split your data file, and run only one regression analysis. Compute a
dummy variable, a variable with values 0 and 1 indicating your two groups.
Compute another variable as the product of this dummy variable with your
independent variable X. Run the regression analysis with three predictors:
your X, the dummy variable, and the product variable. The test of the
regression coefficient of the product variable will tell you whether or not
the slopes of the two regression lines differ. The regression coefficient of
the product variable equals the difference in slopes of the two regression
lines. The regression coefficient of the dummy variable equals the
difference in intercepts of the two regression lines.
This procedure is explained in most books on regression analysis, for
instance Cohen et al. (2003), Applied multiple regression / correlation for
the behavioral sciences. Erlbaum.
===================================
You can't except by the ocular test (eyeballing it). The best bet is to use
the variable that splits the data as a variable in the regression model and
include the interaction of the split variable with the predictor. If the
interaction is significant, then the regression lines have different slopes.
Paul R. Swank, Ph.D.
==================================================
From Ghazi Shukur
You can simply use the so called CAOW or AVONA test. It is sp here:
1- Run a regression on the entire sample (pooled regression) and pick upp
the
Residua sum of squares from this regression and call it RRSS (Restricted
Residual Sum of Squares)
2- Run two seperate regressions on the two sub samples and again pick upp
the
RSS from each of them and call them for RSS1 and RSS2 respectivily.
3. Build an F-test as: F = ((RRSS-URSS)/q) / ((URSS) / N-2k)
where,
F is distributed as F( q, N-2K ), under the null hypotheses of the same
regression line.
RRSS = restricted sum of squared residuals,
URSS = (RSS1 + RSS2) = the unrestricted sum of squared residuals,
N = number of observations,
K = number of parameters, and
q = number of restrictions.
These terms depend on the particular hypothesis being tested. For a test of
overall equation stability, RRSS is from the whole sample period and URSS
is
the sum of squared residuals from each sub-period added together.
========================================================
_________________________________________________________________
Is your PC infected? Get a FREE online computer virus scan from McAfeeŽ
Security. http://clinic.mcafee.com/clinic/ibuy/campaign.asp?cid=3963
