| |
 |
|
|
Science Forum Index » Space - Consult Forum » Binary Logistic Regression...
Page 1 of 1
|
| Author |
Message |
| Anne Harrison... |
 Posted: Tue May 13, 2008 5:35 am |
|
|
|
Guest
|
Hello,
I am conducting binary logistic regression. The results returned to
not include significance, Exp(B), etc. for all the parameters of the
covariate variables. For example, socio-economic status has four
coded quartiles (1-4) however the results include three parameters.
Another variable has five parameters (1-5); the results include four
parameters. Additionally, the results to not logically align with the
coding for the variable (designated 1 in the results is not certain to
be designated 1 in the data set).
How can I force the output to include variables that are being
excluded due to redundancy? Or how can I identify which are being
excluded?
Thanks,
~Anne |
|
|
| Back to top |
|
| Bruce Weaver... |
| Posted: Tue May 13, 2008 11:44 am |
|
|
|
Guest
|
Anne, if posting the same question to multiple newsgroups, please make
one post with the multiple groups listed in the Newsgroups: line.
On May 13, 1:04 pm, Anne Harrison <anne... at (no spam) verizon.net> wrote:
Quote: On May 13, 9:54 am, Mark Ebermann <cfisc... at (no spam) aachener.de> wrote:
One of the values will have been used as a reference category. I think
you need to do some more reading on this.
On Tue, 13 May 2008 08:31:07 -0700 (PDT), Anne Harrison
anne... at (no spam) verizon.net> wrote:
I am conducting binary logistic regression on a data set. The results
do no contain signficance, exp(B) for all parameters of some covariate
variables. For example, socio-economic status has four parameters but
only has results for three. Additionally, it is not apparent that the
number assigned to the parameter in the result corresponds to the
number assigned in coding the variable.
How can I force results on parameters that are dropping out due to
redundancy?
Thanks,
~Anne- Hide quoted text -
- Show quoted text -
Can you explain what you mean by 'reference category', please? I have
been trying to find an answer for some time, looking through many
resources. This is the first that I have been told about reference
category; the best I've found is redundancy.
Thank you, very much.
~Anne
In your example, the categorical variable had 4 levels. Those levels
can be coded with a series of 3 binary indicator variables (sometimes
called dummy variables). E.g., let V be the original variable, and C1-
C3 indicator variables for categories 1-3.
V C1 C2 C3
1 1 0 0
2 0 1 0
3 0 0 1
4 0 0 0
With this coding, the logistic regression would yield coefficients for
C1, C2, and C3. Those coefficients would give you the differences (in
the log-odds of the outcome event happening) for categories 1, 2, and
3 relative to category 4. I.e., category 4 (the one denoted by all
0's) is the reference category.
In the Exp(b) column of your output, you get odds ratios, but again,
they would be relative to category 4.
If you wanted to make category 1 the reference category, you would
omit C1 but include C4, like this:
V C2 C3 C4
1 0 0 0
2 1 0 0
3 0 1 0
4 0 0 1
And so on.
Finally, notice that in SPSS, you do not have to create these sets of
indicator variables. Instead, you can treat your original variable V
as categorical, and choose which of the levels you wish to use as the
reference category. The help files should have good details on how to
do that. (But note that even when you do it this way, SPSS creates a
set of k-1 indicator variables in the background, as described above.)
HTH.
--
Bruce Weaver
bweaver at (no spam) lakeheadu.ca
www.angelfire.com/wv/bwhomedir
"When all else fails, RTFM." |
|
|
| Back to top |
|
| |
|
Page 1 of 1
All times are GMT - 5 Hours
The time now is Mon Oct 13, 2008 2:13 pm
|
|