cancel
Showing results for 
Show  only  | Search instead for 
Did you mean: 
Try the Materials Informatics Toolkit, which is designed to easily handle SMILES data. This and other helpful add-ins are available in the JMP® Marketplace
Choose Language Hide Translation Bar
nbeagan
Level I

3^3 factorial design - Are biological replicates necessary

Hi, I am very new to DoEs and I have a quick question regarding my design and I could really do with some help.

 

Essentially I am looking to optimise my product yield by varying 3 concentrations of DNA additions to cell culture. Ideally, I would like to run the experiment with 3 levels of variation (-1, 0, 1) and with a full factorial design which would be 27 runs and 3 centre point runs = 30 in total. I have run the analysis and the power number is high for all interactions.

 

So essentially my question boils down to this; is there a need to have duplicates for each run if I inlcude duplicate or triplicate centre point runs? I am trying to minimise the number of runs I need. Alternatively, would it be advised to run a 2^2 factorial design with duplicates.

 

Thanks for your help!

3 REPLIES 3

Re: 3^3 factorial design - Are biological replicates necessary

What do you expect the minimum difference in the response to be? What is the typical run-to-run variation in this response? Biological systems are notoriously variable.

statman
Super User

Re: 3^3 factorial design - Are biological replicates necessary

First, welcome to the community.

 

Why center point runs if you have a factorial of 3 factors at 3 levels you will have center point runs.  The question of replication or repetition is important.  How will you estimate the random errors in the experiment space to assess significance?

 

There are many design options.  My advice is to design multiple experiments and compare/contrast each different design and the possible knowledge you can get from each one.  Then consider the cost associated with each design.  Weight the cost/benefit and choose one.  You will likely have to iterate, so don't try to get everything in the first experiment.  KISS (Keep It Simple and Sequential).

"All models are wrong, some are useful" G.E.P. Box
P_Bartell
Level VIII

Re: 3^3 factorial design - Are biological replicates necessary

Do you really need to estimate all the possible interactions in a 3**3 full factorial? Have you taken a look at perhaps an I optimal design that is model specified for all main effects, two factor interactions and quadratic effects? Then if you really need replicates due to the noise as @Mark_Bailey alludes...add them up to your budget allowance?