Regression Error

Discussion in 'SPSS' started by Tan, Oct 20, 2011.

  1. Tan

    Tan Guest

    Hello Everyone,

    I am doing regression analysis and I seem to get this error and unsure
    how to fix it:

    170 proc reg data=fee.women1;
    171 model feevar= age1 race educ service majdep ptsd anx complexity1/
    COLLIN ;
    ERROR: Variable RACE in list does not match type prescribed for this
    NOTE: The previous statement has been deleted.
    172 run;

    WARNING: No variables specified for an SSCP matrix. Execution
    NOTE: PROCEDURE REG used (Total process time):
    real time 0.03 seconds
    cpu time 0.04 seconds

    So my race variable is coded 1,2,3,4,5. Does this mean that my race
    variable have too many levels?

    Tan, Oct 20, 2011
    1. Advertisements

  2. Tan

    Bruce Weaver Guest

    This newsgroup (comp.soft-sys.stat.spss) is concerned with SPSS. "proc
    reg" looks like SAS. Try posting to
    Bruce Weaver, Oct 20, 2011
    1. Advertisements

  3. Tan

    Rich Ulrich Guest

    Like Bruce says, this is not the SAS group.

    On the other hand, the ERROR message is not obscure
    if you take time to read it closely.

    - apparently SAS wants the variables in a regression list to be
    of some "type" other than what RACE is. My guess is that you
    may have called RACE 'nominal' or 'categorical'; and that this
    proc reg is not one that will create dummy variables out of it,
    so it tells you that it is the wrong type.

    Having to label your variables with a TYPE may seem like a
    pain-in-the-butt if you always know what you are doing,
    and never make a mistake. In this instance, it looks like it
    prevented an erroneous analysis ... unless you happen to
    have codes for RACE that do represent some increasing/
    decreasing amount of some quality.

    So, if I am not mistaken, you should replace RACE with
    several dummy variables, or else, change its type.
    - and this NOTE tells you that your request has been deleted,
    so nothing remains to RUN.
    Rich Ulrich, Oct 21, 2011
    1. Advertisements

Ask a Question

Want to reply to this thread or ask your own question?

You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.