Hi,
I am a statistician currently working on causal
analysis in non-orthogonal Anova.
The approach I follow is similar to the one of Rubin and
Rosenbaum (propensity scores), but is more oriented towards
confounding variables.
In particular I have developped a procedure to test causal
effects in non-orthogonal Anovas in which some of the factors
may be regarded as treatments and the others are mostly
intrinsic properties of the units and may be in some cases
regarded as potential confounders.
A typical example is a study to investigate the effects of
three different therapies (first factor) on a continuous
response variable Y (greater values of which reflect an
improvement in patients conditions). As second factor I
consider the urge for a therapy (measured on a scale from
1=dramatic to 3=negligible).
I would be very grateful for any comment or hint on
propensity scores-based model or on similar approaches.
I am particularly interested in knowing of situations or
data for which this kind of approach may be of interest.
Thanks in advance
Olivia
-------------------------------------
Dipl.-Math. Olivia Wuethrich-Martone
Institut fuer Psychologie
Friedrich-Schiller-Universitaet-Jena
Am Steiger 3/Haus 1
D-07743 Jena
Tel.: +49-3641-945236
E-Mail: [log in to unmask]
-------------------------------------------------------------
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|