Dear all,
I have recently been working on a logistic model
with 18 binary explanatory variables.
I split these 18 variables into 4 groups and drew
up a conceptual hierarchical framework as done by
Victora et al (1997).
I then used the changes in likelihood to assess how
much of the effect of each group of variables was
mediated by groups further down the pathway.
I have two questions.
1) Is this a reasonable approach - I didn't fancy
individually examining lots of odds ratios at each
step.
2) I have now moved onto a 3-level polytomous regression
and it has become much more difficult to interpret these
results - any suggestions? Is this no longer of any use?
thanks
Jon
Victora, C.G., Huttly, S.R., Fuchs, S.C., and Olinto, M.T.A. (1997)
The role of conceptual frameworks in epidemiological analysis:
a hierarchical approach. International Journal of Epidemiology 26 (1)
|