| Philippe Hespel 2005-04-26, 4:06 am |
| Hi all:
I am analyzing data Y=F(A,B,C,D,...,K,L,M,..)
I have one dependent variable Y, which can be 0 or 1 (e.g. failure or
success).
I have several independent variables, which can be split up into 2
categories:
(1) discreet variables (A,B,C,D,...) which can only have 2 or 3
different values
(2) continuous variables (K,L,M,..)
I do not know the relationship Y=F(A,B,C,D,...,K,L,M,..), but I have
several sets of observations of A,B,C,D,...,K,L,M,..
MY QUESTIONS ARE:
How can I solve following problems (which instructions,..):
(a) which variable(s) has/have the strongest predictive power?
(b) how to define a relationship/model between independent and
dependent variables when I have no knowledge about characteristics?
(c) if model is defined, how much variation is explained by model?
(d) if I have a new set of independent variables how can I predict
the dependent variable Y, which is discreet (e.g. giving these
variables A,B,C,...,K,L,... what is the probability of getting
success or Y=1)?
MANY THANKS for any help!!
Philippe
|