Hi all,

With my dataset I computed relations between my dependent variable and
several independent variables with spearman correlations and binary
logistic regression.

These two methods show different correlations between the variables.
On which should one base conclusions, regression or spearman?

Thanks

Play around with google and those two terms for more information. 
There are several good websites that talk about these statistics in
depth.  The answer depends on your research question.  At base
remember that a correlation is just the strenght of the relationship
between two variable (i.e. the correlation between height and weight
is pretty good); however, correlation DOES NOT imply causation.  While
it examples like height and weight certainly imply that the taller you
are may impact your weight, that certainly does not hold true for all
correlations.  For example, ice cream sales were found to be highly
correlated with shark attacks, but I don't think anyone would suggests
that sharks track ice cream sales and plan their attacks according. 
Just because two things correlate doesn't mean that they are really
related in any way.  A logistic regression is used when you want to
predict a dependent variable.  In essence, to say that ice cream sales
really are related to shark attacks and can predit their occurrance. 
My guess, since you are using multiple IVs is that you probably need a
regression and not a correlation, but without knowing more about your
research question, its impossible to say.  Good luck!!

Hi!

thank you very much for your help. Well, i am doing research on the
impact of values on the adoption of an innovation. I got about 100
people to fill in my questionnaire (which consists of a  worldwide
value survey developed by Shalom
Schwartz(which measures the scores on 10 value domains) and the
question wether or not the respondent did or did not adopt the
innovation plus several demographical questions.)The 10 vales are all
ordinal variables and the adoption variable is a nominal one (yes/no).

I computed relations for the 10 values and adoption with spearman
which showed some correlations which in my view do indicate causality
since a "-" is placed before the coeffecient if it is a negative
relationship.

I also computed binary logistic regression with the adoption variable
as the dependent variable. These showed some different correlation
between the 10 values and "adoption" than with spearman which confused
me.

How do yo know when you have the best model with binary logistics?
when the predictability of the model is the highest?

I hope you can help me a little bit more. if so i will defenitely
raise the list price accordingly.

Thanks

Thijs