Dear Allstaters:
I would like to know your opinions ( I am not an expert in statistics) about
- the use of a specific test to compare regression slopes or a more
general test
- to work with original data or with means that have been obtained from
groups of that set of data
I need to compare the regression slopes of the two straight lines
L1 and L2 (y=a+bt)
that I fit with a set of values (L1 to m=0 and L2 to m=1) where
y: continuous response variable
t: discrete time (t=1,2,3,4,5)
m: qualitative variable (m=0,1),
In each regression analyses, for each value of t I have 8, 9 or 10 values
for the y response variable.
1)What do you think about to work with the original data or with the
means of each group of values of t to do these fits?
(Really the variability for each set of data corresponding to a specific t
value is important, so the R-square coefficients in the ANOVA tables
are small when the original data is used).
2)A reference: Models in Biology (Brown and Rothery) points out that the
comparison of regression slopes can be tested using an approximate
t-test. The test-statistic is the difference of the estimated slopes divided
by the standard error of the difference. Also, they explain that the degrees
of freedom of the t-statistic can be calculated by a rather complex
formula.
Do you know any reference where I can find this formula?
Does this test (unknown name?) appear in statistical packages
like SAS or SPSS?
3)On the other hand, do you think that would be better to use the
covariance analyses to deal with this problem?
Would you use the original data or the means of the grouped data in
each t to carry out the analyses?
4)In general, do you know any reference about the use of the means
of grouped data like entrance data (instead of the original data) to be
analysed with ANOVA?
Sorry for the extension of this e-mail, but I would be very thankful for
any help, suggestion or reference.
In advance, thanks for your time.
Marta
Marta Ginovart
ESAB, UPC
Urgell 187, 08036-Barcelona, SPAIN
Tel (343) 430 4207 - FAX (343) 419 2601
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|