Note as well that if you use GLM->Univariate instead of
REGRESSION, you can enter categorical variables as Factors,
and SPSS will do the dummy coding for you. If you check the
Parameter box under Options, you will get a table of
parameters (i.e., regression coefficients). The GLM method
does not provide all of the same features that are available
via REGRESSION, however (e.g., change in R-squared,
> Judd & McClelland's book is all about this:
I think you should read the Cohen & Cohen book (Applied Multiple
Regression/ Correlation Analysis for the Behavi*Sciences) as
part of a graduate education in methods. Cohen was one of the folks
who introduced MR as a general tool for ANOVA.
The place that I have seen the most stuff about diagnostic tests
has been in econometrics. Even though a lot of their problems
that need diagnosing are intercorrelations or non-homogeneity
that arise most seriously in time-series data, it probably won't
be a waste of time to browse in a book or two, such as,
Econometric Analysis by William H. Greene.
"Taxes are the price we pay for civilization." Justice Holmes.
Also, don't hesitate contacting me directly, I use these procedures
routinely. Perhaps I can help.
> > > Hi,
> > > I am currently working on my dissertation for my Masters of Philosophy in
> > > Methods of social science research, I am trying to construct a regression model
> > > but the data set I am using contains mostly ordinal variables. I have created
> > > dummy variables for most of these but am unsure about interpreting the
> > > co-efficients and how to include, and also how to conduct diagnostic checks
> > > (for heteroscedasticity, multi-collinearity, omitted variables, and so on). If
> > > anyone konws of a helpful book or any other information to do with dummy
> > > variables please contact me.
I may be mistaken, but I thought the original poster meant
that a lot of the predictor variables were
categorical/ordinal, not the outcome variable.