--- Marc Campo <[email protected]> wrote:
> I am entering ordinal variables into LR as
> continuous I have 5
> ordinal indep vars all based on frequencies coded:
> 0 - 0x.day
> 1-1-5/day
> 2-6-10x/day
> 3 - 11-15x/day
> 4 - More than 15
>
> Three of them are highly related and I would like to
> combine them. However, just adding them establishes a
> new variable that is not ordinal. What to do?
Marc:
Adding the variables is equivalent to constraining the parameters of
the three variables to be the same. Call the three variables you wanted
to combine x1 x2 x3 and the dependent variable y. The unconstained
regression would look like:
y = b0 + b1 x1 + b2 x2 + b3 x3 + other vars
If the regression parameters are constrained to be equal, then the
regression equation would look like:
y = b0 + b x1 + b x2 + b x3 + other vars
which can be simplified to:
y = b0 + b (x1 + x2 + x3) + other vars
Constraining the regression parameters to be equal seems a reasonable
way of combining variables that have the same scale, as is the case in
your problem. Additionally, the regression parameter will maintain the
original scale: the effect of one step increase on your ordinal scale
on your dependent variable. Note that your original scale wasn't
strictly ordinal to begin with: you already assumed that each one step
increase has the same effect, i.e. you already made an assumption about
the distances between the categories by adding them as continuous
variables.
HTH,
Maarten
-----------------------------------------
Maarten L. Buis
Department of Social Research Methodology
Vrije Universiteit Amsterdam
Boelelaan 1081
1081 HV Amsterdam
The Netherlands
visiting address:
Buitenveldertselaan 3 (Metropolitan), room Z434
+31 20 5986715
http://home.fsw.vu.nl/m.buis/
-----------------------------------------
Send instant messages to your online friends http://uk.messenger.yahoo.com
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
