Skip to content

Experimental Design

2 messages · Isaac Neuhaus, Spencer Graves

#
I don't know if this is the best place to post this question but I will 
try anyway. I have two experiements for which I use one-way 
matched-randomized ANOVA for the analysis and I would like to compare 
different treatments in the two experiments. The only common group in 
the two experiments are the controls. Is there any  ANOVA design that 
allows me  to  make this comparison taking into consideration the 
confounding effect? Any help would be greatly appreciated.

Isaac

A representation of the experiments follows:

Experiment 1
           Control1     Treat1      Treat2
Blk1          s1          s2          s3
Blk2          s4          s5          s6
Blk3          s7          s8          s9


Experiment 2
           Control2     Treat3      Treat4
Blk1          s1a          s2a          s3a
Blk2          s4a          s5a          s6a
Blk3          s7a          s8a          s9a

Control1 and Control2 I are the same control cell line. I would like to 
compare Treat1 to Treat3 and Treat 4 and also I would like to compare 
Treat2 to Treat3 and Treat4. The fact that those experiments are done in 
two different blocks will confound the interpretation. Can I use the 
common control group to build a model? Should I include one of the 
treatments in future experiments to test my model?
#
It looks to me like you have two blocking variables with 1 control 
group and 4 treatment groups, with the control replicated between the 
"master blocking variable" = "experiment 1 vs. 2".  (The minor blocking 
variable occurs at 6 levels unless "Blk1" in Experiment 1 somehow 
relates to "Blk1" in Experiment 2.)  People who deal with this routinely 
could probably provide R code plus citations to the literature where 
this kind of analysis is discussed.  I would write an appropriate model 
and do the analysis.

	  And yes, I would want to confirm any encouraging results in a future 
experiment.

hth.  spencer graves
Isaac Neuhaus wrote: