Date: Tue, 1 Jun 2004 20:14:35 +0100
Reply-To: juan carlos <bello33uk@YAHOO.CO.UK>
Sender: "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From: juan carlos <bello33uk@YAHOO.CO.UK>
Subject: BOX TIDWELL
Content-Type: text/plain; charset=iso-8859-1
I want to test for linearity between continuous
independent variables and the dependent variable in a
logistic model (which contains a group of variables of
different scales) using what is known as Box-Tidwell.
What I understand is to have the product of each
independent variable and its log and to add this
product to the model to check its significance.
However, I am not sure of the following points:
If I have more than one continuous variable, shall I
do that with them all and inter the products (terms)
in one time, so that I will have a model with these
continuous variables and their products with their
logs. Or shall I do that separately for each variable?
I think the significance will be different? And in
case you have four or five independent variable of
interval scale, you have different possible models.
Then shall you test this linearity with all this
models? Help me please
Yahoo! Messenger - Communicate instantly..."Ping"
your friends today! Download Messenger Now