[R-sig-ME] Issue with experimental design?

[David Winsemius]

On Oct 12, 2013, at 4:11 AM, Yla Savh wrote:

> Dear forum members, Could you please help me confirm if there is a problem with
> the experimental design? There are two initial groups (vitamin D deficient and
> normal group). Then, the D deficient group has two treatments (with a low D dosage
> and a high D dosage), while the normal group has only one treatment (maintenance
> therapy). 
> Initially, I thought it might be a nested design (treatment
> nested within groups, where the treatment is a fixed effect, and groups and treatment
> nested in groups are random effects. However, I do not think it is a correct
> design as the groups did not include the same treatments. Am I correct?
> I see only one solution where we will have only one or two
> groups and the same treatments should apply to each of them. Are there other
> solutions?Thanks,Julia
> 

Perhaps a different opinion: You are really only "experimenting" with the "vitamin D deficient" group. There is no design basis for including the "normal" group in any inferential analysis. I suppose you can construct plots to look at descriptive comparisons but I see little valididity to including the structurally different (different by design) normal group in any analysis where you would be constructing effect measures. I suppose it might give you some estimate of variability and you would be able to describe differences, but you wouldn't have a solid basis for making inferences in that analysis because the range of initial vitamin-D values would be of necessity disjoint.


> 	[[alternative HTML version deleted]]
> 
> _______________________________________________
> R-sig-mixed-models at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models

David Winsemius
Alameda, CA, USA