Logistic Regression, Dummy Variables

[zstatman]

Running a BLR with the DV coded as Yes/No based on whether or not the
respondent completed a course

Have several Dummy IVs but looking at 1 pair coded from another variable
based on participation in a Learning Center program.

So then under Dummy_1 there are, e.g., 50 codes as a 1 (or Yes they did)
with the remaining cells being a 0 then just the opposite for Dummy_2. The
Dummy pairs were created using the Create Dummy Variables extension under
V18.

The issue is the following "error" message that I can't understand:

"Due to redundancies, degrees of freedom have been reduced for one or more
variables."

Thoughts appreciated

WMB

=====================
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

[Bruce Weaver]

Statmanz wrote

Running a BLR with the DV coded as Yes/No based on whether or not the
respondent completed a course

Have several Dummy IVs but looking at 1 pair coded from another variable
based on participation in a Learning Center program.

So then under Dummy_1 there are, e.g., 50 codes as a 1 (or Yes they did)
with the remaining cells being a 0 then just the opposite for Dummy_2. The
Dummy pairs were created using the Create Dummy Variables extension under
V18.

The issue is the following "error" message that I can't understand:

"Due to redundancies, degrees of freedom have been reduced for one or more
variables."

Thoughts appreciated

WMB

It  sounds as if you might be entering both of the dummy variables in the model.  But because Dummy_2 contains exactly the same info as Dummy_1 (with 0s and 1s swapped), you must include only one of them.  If you include Dummy_1, the reference category for the odds ratio will be those subjects who have Dummy_1 = 0.  If you include Dummy_2 instead, the reference category will be those subjects who have Dummy_2 = 0.

Extending this, if the original categorical variable has k categories, you must include k-1 dummy variables.  The kth one is completely determined after that.