iSixSigma

When to Use Center Points

Six Sigma – iSixSigma Forums General Forums Methodology When to Use Center Points

Viewing 5 posts - 1 through 5 (of 5 total)
  • Author
    Posts
  • #55826

    Armando Flores
    Participant

    First of all, center points are used to detect curvature, basically this are points set halfway in a DOE, if in case Ct Point are significant then you need to use Surface Response method to analyze data, but i still have some questions.

    1. What if i design an experiment without center point? what would be the consequences of this. Lets say Im happy with the results and i didn’t use ct points. Am i wrong?

    2. Should i always use Ctp points to design a DOE to see if curvature exist?

    0
    #201839

    Robert Butler
    Participant

    First off – they are used for more than just a check for the existence of some kinds of curvature. If you have your design and have not included replicates then running a center point with a single replicate not only buys a check for possible curvature it also gives you an economical measure of pure error (we are, of course, assuming variables that are either interval or ordinal in nature).

    1. It depends on the kind of design. If it is strictly a 2 level factorial then you won’t be able to check for curvature. If it is a design that has variables that are nominal then you won’t be able to run a center point.

    2. Assuming you can run a center point with the variable you have then the issue becomes one of time and money.
    a. If you have no reason to expect curvilinear behavior over the regions of interest and the cost per experiment is high then you may not have the option of adding more experiments.
    b. If you can build and run a center point then – why not? As noted in the opening statement an unreplicated design with center point replication gives you a lot of information for a minimum of cost.

    0
    #201840

    Armando Flores
    Participant

    Robert perfect response, yes we can use ctp point for an alternative way to increase power and no need to use more replicates. also point 2 is clear for me, thanks

    what i didn’t understand is this;
    if it is strictly a 2 level factorial then you won’t be able to check for curvature
    Even a 2^2 design you can add center points so you can be able to check curvature
    what am i missing.?

    0
    #201841

    Robert Butler
    Participant

    If it is strictly a 2 level factorial design then of course you can’t check for curvature. A straight line is defined by 2 points whereas a curved line requires a minimum of 3. A strictly 2 level factorial has only a min and a max setting for each of the variables – 2 points so, only a straight line and interactions expressed as non-parallel straight lines.

    1
    #201842

    Armando Flores
    Participant

    Thanks Sr.

    0
Viewing 5 posts - 1 through 5 (of 5 total)

You must be logged in to reply to this topic.