iSixSigma

DOE – Replicates or Repeats?

Six Sigma – iSixSigma Forums Old Forums General DOE – Replicates or Repeats?

Viewing 6 posts - 1 through 6 (of 6 total)
  • Author
    Posts
  • #37782

    Thaly
    Member

    I’m performing a DOE and I’m unsure on deciding whether my runs as such are replicates or repeats. The parts on which the response is meaasured is coming out of a forge press line (response is straightness of the features in the part). So if I take three successive samples out of the product line for each treatment(setting the appropriate levels for each factor, of course), would that constitute a three replicate DOE or a 1 replicate, 3 repeat experiment?
    The noise in the process is comparatively high and so I know I cannot get away with just one replicate.
    If I can consider successive parts out of the forge line as repeats, it’d be a lot easier for me. (I realize I’m not considering randomization, but I’m confident of getting normal data or transforming it in the extreme case).
    Thanks in advance for suggestions.

    0
    #111972

    Darth
    Participant

    Replicates are preferred over repeats and captures the variation of the process.  That usually entails setting up the process with the same conditions later on as determined by the randomization of the runs.  Repeats are obviously easier to obtain but don’t quite give you the same amount of information needed to assess the variation of the process.  The randomization of the runs also allows for some blocking out of lurking variables that you are not interested in testing.

    0
    #111975

    VILLAMIL
    Member

    Usually, when there are 2 factors to DOE(A, B), it is repeat if someone choose 2 successive samples in A condition 1, B condition 1.
     
    If someone choose 1 sample in A condition 1, B condition 1 and reset the factors, choose another sample.
    That is replicate.
     
    Your case is exactly repeat.
     
    However, when it is repeat, decomposition of experimental error and observation error is impossible if there is no replicate, and block’s effect can¡¯t be verified.
     When collect samples if there are a lot of Noises, it is good that use ANOCOVA.(If you want to use this method, you must collect data for noise valuable. If randomization or setting of equipment makes getting successive samples difficult, it is good method, SPD.

    0
    #112273

    Jonathon Andell
    Participant

    You have gotten good answers. To respond directly to your question, a repeat is when you keep the same run settings and sample the process several times. A replicate means that you return to the run settings later in the experiment. As Darth said, that is more complex, but also richer in data. If cost is no object you have the option to do both, but that seldom is practical.

    0
    #112842

    ALEK DE
    Participant

    Hi Thaly ,
    Use of  repeats or replicates depends on the the type of variation you want to addrss.Replicates are from distinct experimental runs, spread over a longer period of time. In repeat , you take successive measurements without disturbing the set up. This more of a short term variation in the process. Replicate can include sources of variation which are not included in repeat measurements. For example, replicates can include variation due to change in equipment setting  or operators or place of manufacturing or any  other factors that may change over time. Replicate measurements can be more expensive and time-consuming to collect.
     
    Thanks
    Alek
     

    0
    #112843

    ALEK DE
    Participant

    Hi Thaly ,
    Use of  repeats or replicates depends on the the type of variation you want to addrss.Replicates are from distinct experimental runs, spread over a longer period of time. In repeat , you take successive measurements without disturbing the set up. This more of a short term variation in the process. Replicate can include sources of variation which are not included in repeat measurements. For example, replicates can include variation due to change in equipment setting  or operators or place of manufacturing or any  other factors that may change over time. Replicate measurements can be more expensive and time-consuming to collect.
     
    Thanks
    Alek
     

    0
Viewing 6 posts - 1 through 6 (of 6 total)

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