DOE Question

Six Sigma – iSixSigma Forums Old Forums General DOE Question

Viewing 5 posts - 1 through 5 (of 5 total)
  • Author
  • #46710


    Hello All,
    I am looking at some software options for DOE.  Some can handle many factors, many replications, many blocks, etc.  My question is not really about the software but about common practice.  For those of you who performance experimental designs on a regular basis, please let me know, based on your experience, the following:
    1. How many factors do you normally  include in a DOE?  What is the max you have used?
    2. How many replications do you typically use?  What is the max you have used?
    3. What percentage of your DOEs have involved more than one block?  What is the max you have used?
    4. What is maximum number of runs you have ever used in a DOE?
    I understand that these questions depend on the what you are investigating.   I am looking for some typical DOEs plus the extremes.
    Thanks for the feedback.


    DOE student

    Bill: you will “acquire” answers to these questions very soon as you start plying around with some experiments. Some do classical doe (very statistical) while others can look at more practical doe like the taguchi method. Eitherway the point is to do a good design of experiment. Choice and RESOURCES will determine most of your questions. Have fundoe student 



    If you’re serious about DOE, you might want to do some studying beyond the  Certified/ Trained Six Sigma Black Belt DVD and comic book level. Try the following:
    Design and Analysis of Experiments. D.C Montgomery, Wiley
    Statistics for Experimenters.Box, Hunter and Hunter, Wiley
    Applied Linear Statistical Models. Neter, Li, et al. McGraw – Hill (I think)



    Thanks for answering.  I have gone beyond the comic books. I have read Box, Hunter and Hunter in detail.  I have Montgomery’s book.  All I wanted from my questions was some answers from people who have done quite a few experimental designs. 


    Robert Butler

     I’ve been setting up and running DOE’s for about 25 years.  These have been in the fields of plastics (both manufacture and use), aerospace, specialty chemicals, and medicine. 
     1. Designs involving 10 to 15 factors have been quite common.  The number of measured outputs varies but 5 to 10 measured responses per experiment would not be out of the ordinary.  As for the maximum – the biggest design effort I was ever involved in had 151 variables.
    2. The only time I’ve ever run a complete replication of a design was when I was ordered to do so – and this has only happend a few times in my life.  The whole point of a design is economy of effort and full replication is a waste of time and money. If all of the variables are continuous I will confine replication to a couple of center points.  If it is a mix of type and continuous I will pick a couple of points at random for replication (with the single exception noted below).
    3. Offhand I’d guess about 30%.  Within this group 2-3 blocks is probably typical.  In almost every instance the blocking was of the split plot variety – namely the blocking was due to the difficulty of randomizing one or more variables in the design (e.g. changing a temperature profile in an extruder, changing the feedstock from different suppliers, etc.) In these instances replication involves the partial replication of one of the blocks. In most instances the partial block will have more than two experiments.
      4. As mentioned above 151 was the max for an effort but these were not checked in a single design.  I’ve run several designs with 40-60 experiments each but I can’t be more specific than that.  For me, designs with 12-36 run are quite common.

Viewing 5 posts - 1 through 5 (of 5 total)

The forum ‘General’ is closed to new topics and replies.