Read up on regression.
Regression coefficients are sometimes called "partial regression
coefficients". Putting in the word "partial" serves to call attention
to the fact that every user should stay highly aware of: that the
"effect" of each variable is measured *after* the presence of
every other variable in the list. That is why two rather similar
variables will share an effect, or one might "knock out" the other.
(Or, the combined contribution might even show up as looking at
the difference between the two, which is a more complicated matter.)
That is why you don't want to put into the set of predictors variables
that are redundant, even approximately: You do not get an intelligible
model as the result unless you are willing and able to understand
the way the parts work together.
I think another program explicitly uses the word "redundant" for
the situation that you describe. Whether you use that word or not,
redundancy is what you have. Ask yourself -- If I have a variable
Male/Female, What is the contribution *added* by another variable,
Female/Male? What should the program tell you? - Wisely (IMO),
SPSS chooses to suppress such variables... as you see.
Creating too many dummy variables for the available degrees of freedom
is one way to create redundancy. The other most popular way is to
try to include a Total along with all the items that make it up, or a
Grand Total along with the Totals that go into it.
Read up on regression.
--
Rich Ulrich
> Date: Fri, 5 Jun 2015 18:59:41 -0700
> From:
[hidden email]> Subject: Full model instead of a reduced model in regression
> To:
[hidden email]>
> Hi, I am running a multinomial regression with about 12 factors and the final
> "reduced" model supresses a few of my variables with the message "This
> reduced model is equivalent to the final model because omitting the effect
> does not increase the degrees of freedom". and hence in the final table a
> missing value is set for their odds ratios, P value ect. Is there a way to
> not reduce the model so i can actually see the odds ratio and P value of all
> my factors?
> thanks.
>
=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD
Locked
