At 06:44 AM 4/22/2014, dimitrios wrote:
>When I use age as a continuous variable, it is not linear and
>therefore I cannot use it.
>If I use age as ordinal (18-40, 41-60, 61-80), I guess I do not need
>to worry about linearity.
>Results come back similar and from practical point of view, it does
>not change a lot.
I refer you to Bruce Weaver's remarks, which are absolutely right.
Let me add a couple of points:
. If you're getting about the same results with age as continuous
variable as you do when dividing age into categories, then perhaps
the effect of age *is* linear, or pretty close to linear, and you
should use it that way in your model.
. If you suspect a non-linear effect of age, a good step is to then
add an age-squared term to your model. I would make it the square of
the difference of age from some value near the center of your range,
rather than the square of age itself. (I advocated for this in this
thread a while back; you can read the back postings, for the
arguments for and against it.)
. Finally, a point that I don't think has been made: If you break a
variable into (n) categories, then some procedures (I don't know if
it's true of the one you're using) effectively add (n-1) parameters
to the model. More parameters means you need a larger sample; having
many categories in the model can create difficulties if your sample
size is modest. Note that a linear and a quadratic term in age are
just two parameters, together.
=====================
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
