Logistic regression controlling for variables vs. layer in crosstabs

#1
Hi everyone,

I would like to know how SPSS controls for variables in logistic regression in order to report the correct "controlled" OR.


The specific case which bothers my is as follows:
Lets say I have one IV tumor size which is either small tumor size or large tumor size the other IV is Risk classification with low risk and high risk.
My dependent variable is complications yes or no.
Unfortunately people with a high risk often also have a large tumor.

Now to assess the actual effect of the risk classification I used the layer tumor size and did a crosstab which gave me the OR for high vs low risk patients when they have large tumors and high vs low risk patients when they have small tumors.
I assumed this was the controlling of variables.

Subsequently I heard that, when one performs a logistic regression in spss under table "variables in the equation" the OR (Exp(B)) for each variable while controlling for the other variables is displayed.
Unfortunately these ORs do not match with the ORs of the corsstab version.
What is the difference in that case?

Thanks a lot, I appreciate your help!
 
Last edited:

Karabiner

TS Contributor
#4
What I am asking myself, is it nice of someone to post the
very same question at the same time into different help
forums (say, for example talkstats and CrossValidated),
so that people who like to be helpful might work on the
same question in parallel, but without knowing each other?

Ok, someone doing so might be a very important person,
and/or might just have fun in wasting other peoples time,
effort & goodwill, but still...

(Just asking.)
 
#5
I'd say it just displays other people’s urgency.
And if you'd just have the capacity or the gratefulness of replying with an appropriate statistical answer once, than you'd not waste any time giving unqualified answers.
Also any unanswered post will be deleted or even better will display the answer as soon as any of the different forums finds an answer. This is redundant and even helps people with the same question. However when you experience that many questions aren’t answered at all in a forum than you might consider to maximize your efficacy.
Can you live with that or do you wanna turn this platform into a philosophical discussion forum?
 
#6
This is what he said:

Hi everyone,

I would like to know how SPSS controls for variables in logistic regression in order to report the correct "controlled" OR.


The specific case which bothers my is as follows:
Lets say I have one IV tumor size which is either small tumor size or large tumor size the other IV is Risk classification with low risk and high risk.
My dependent variable is complications yes or no.
Unfortunately people with a high risk often also have a large tumor.

Now to assess the actual effect of the risk classification I used the layer tumor size and did a crosstab which gave me the OR for high vs low risk patients when they have large tumors and high vs low risk patients when they have small tumors.
I assumed this was the controlling of variables.

Subsequently I heard that, when one performs a logistic regression in spss under table "variables in the equation" the OR (Exp(B)) for each variable while controlling for the other variables is displayed.
Unfortunately these ORs do not match with the ORs of the corsstab version.
What is the difference in that case?

Thanks a lot, I appreciate your help!
yes only two groups
I'd say it just displays other people’s urgency.
And if you'd just have the capacity or the gratefulness of replying with an appropriate statistical answer once, than you'd not waste any time giving unqualified answers.
Also any unanswered post will be deleted or even better will display the answer as soon as any of the different forums finds an answer. This is redundant and even helps people with the same question. However when you experience that many questions aren’t answered at all in a forum than you might consider to maximize your efficacy.
Can you live with that or do you wanna turn this platform into a philosophical discussion forum?
 
#7
What I am asking myself, is it nice of someone to post the
very same question at the same time into different help
forums....

(Just asking.)
I saw that double post too. I find it not very polite!

But Mo V., Karabiner asked you a question! If you answer that, maybe he will answer you original question?

(And I am sure that Karabiner has "the capacity" but now maybe not "the gratefulness" to answer.)

Everybody should know that nobody here has any duty at all to answer any question.

- - -

The original question was written a little bit fuzzy. For other readers here: If your question has not been understood and therefore not been answered, how clever is it then to write exactly the same formulation once again (although it was on an other forum)?
 
#8
I had no intention or whatsoever for being rude, mean, not nice or not polite!
I beg your pardon if you for yourself interpret it that way.
Looking at cause and effect I cannot recon any expenses for you and I apologize for you getting an email notification of a new post twice. Even the time it takes to realize that the second post is the same as the first one you just read, does not make me feel like I belong into the category not nice.
I could imagine that it is morally a bit offensive with regards to the creators of these forms if someone uses the “competing” web page for the same purpose as well.
Maybe I can be more repentant if you can explain to me what harm I caused to anybody, or why exactly I would be mean.

Thanks so far.

FYI I personally felt a little disturbed to read nothing but a whole post of quotes of my above mentioned. Sorry but I could not see the sense in that first pods of you…
 
#9
Also since you find my first post a bit fuzzy this might be due to the fact of not being a statistics guru and not finding any good answers for many SPSS issues online or in the users guide. However if you tell me what the obstacle is in order to understand my original question better. I'd be very pleased to re formulate.