We have an ANOVA model (unbalanced data design) which shows certain
main effects to be significant. However, when we seek to isolate the nature
of the effects using multiple comparison methods (e.g. Tukey's method)
none of the pairwise differences are significant.
I am aware that the results of multiple comparison methods can difficult to
interpret, especially when using unbalanced data, so what I am looking for is
a good source of guidance regarding the best practice in this area.
TIA.
--
John Polak, Head of Transport Section,
Department of Civil Engineering
Imperial College
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|