iSixSigma

DOE…How to get the data

Six Sigma – iSixSigma Forums Old Forums General DOE…How to get the data

Viewing 8 posts - 1 through 8 (of 8 total)
  • Author
    Posts
  • #49318

    Jose Ramon
    Participant

    Hello eveyone.
    I have a question about DOE.
    But let’s us make it with a case:  les’s say that we have a drugstore and we want to analyze the impact of customers in the morning , in the afternoon and the speed time ( service time) in the drugstore:
    A patients in the morning
    B patients in the afternoon
    C speed time
    Experiment full factorial 3 X 3 = 9 runs
    A       B      C
    1       -1       -1
    1        1       -1
    1        -1       1
    1        1         1
    If the response that I am analyzing ( for instance : waiting time).  How can I collect this data.  in the first run A = 1  B = -1 and C = -1…..does this mean that I have to collect the data for this run when the customers in morning are a lot + the customers in afternoon are few ( But how can I know this ?…..based on sales ? or number of prescriptions?) + when the cashier takes a little bit time to serve a customer?
    This means that I have to do the same for all the runs? or all the data is collected in the morning and in the afternoon for all Xs ?
    Please let me know If I am wrong
    Thanks so much
     
     
     

    0
    #168413

    Chris Seider
    Participant

    Jose,
    If you have 3 factors and you are running a factorial design with no repeats and replicates, you will have 2^3 runs or 8–not 9.  It is not clear to me about your factors but I don’t think your design is good since patients in the morning and patients cannot simultaneously go through your process–which is required for factors in an experimental design.
    Consider this example.  If you identify waiting time as your output, then set your inputs at the low and high (+ and -) level following Yates algorithm or randomize if you can.  Your inputs could be number of patients, speed time (I don’t know what this term means), and another factor.  You will end up setting your factors at the 8 different combinations and see how your process performs (after setting the factors to one of the combinations, record the output and then change to another combination of conditions for your factors).  One of the combinations will have all 3 factors set at the low level while another will have them all at the high level and all the 6 other combinations.

    0
    #168418

    Joe BB
    Participant

    Ramon,I do not think that your problem has been clearly defined. You gave the input factors but you did not define what your output is and 3×3=9 does not mean much. You define your design in terms of full or fractional factorial and in terms of how many factors are involved and how many levels. If you have 3 factors using 3 levels, you will not end up with 9 runs and if you are using a 3 factors 2 level design, you would end up with 8 runs, not 9.Please redefine your problem statement and the factors involved and we might be able to help you.

    0
    #168455

    Robert Butler
    Participant

    You have posted this same question on this forum and on the healthcare forum numerous times over the last several months. With each posting you have received comments and thoughts from a number of people (including myself) and each time your follow on postings gave the impression that you had given little or no thought to what anyone said.  You leave the reader with two distinct impressions:
     
    1.      You are bound and determined to run an experimental design regardless of its appropriateness to your problem.
    2.       You don’t know what a design is or what it is supposed to do but you are determined to find someone to put the stamp of approval on whatever it is that you are calling a design.
     
    I’m not trying to be insulting and I’m not trying to discourage you from posting questions but I think you need to sit down and actually read the posts everyone has offered on your behalf. I also think you need to get a basic book on design and understand what they do and why, at this stage of your effort, they are not appropriate.
     
    Let’s rephrase your posts.
     
    What you are asking is the following:
     
    A) Given a random draw of customers with a variety of problems and some kind of service set up what parts of the service process contribute the most to customer waiting time?
     
    B) Is this waiting time a function of the time of day when the customer arrives?
     
    C) Is this waiting time a function of the number of people that show up at a particular time of the day?
     
     The fourth part, which you haven’t mentioned, should be – is there a relationship between the wait time and the kinds of problems the customers bring for me to solve?
     
     
    The best advice that has been offered with respect to answering these questions are the posts by Stan and Bob when you posted the question over on the healthcare forum. 
     
    http://healthcare.isixsigma.com/forum/showmessage.asp?messageID=3794
     
    http://healthcare.isixsigma.com/forum/showmessage.asp?messageID=3798
     
     
    Everything I’ve read (and I’ve read all of your posts and the responses) argues for just what these two suggested.
     
    Once you have walked the process and defined it in detail (block diagrams, fishbone diagrams, etc.) and looked at the issues of queuing you should be able to identify those parts of the process that are contributing the most to waiting time
     
    A)    Overall
    B)     As a function of customer problem complexity
    C)    As a function of time of day (if any)
    D)  As a function of number of customers per unit time (if any)
     
    Armed with this information you can start to think about ways to change/modify the process to significantly reduce customer wait time.

    0
    #168456

    BTDT
    Participant

    Robert:You saved me trying to compose a diplomatic response. Thank you for saying what I was thinking and setting the poster in the right direction.I have run across more than a few people who were so impressed with the power of DOE, that they were determined to find a way to use one in their projects. This sounds like one of these “solution in search of a problem” situations.Cheers, Alastair

    0
    #168471

    Jose
    Participant

    Thanks a lot Seider, Stan , Robert,and Bob.
    I agree with Mr. Bob about first study your gemba before doing something. “We cannot improve what we do not know” and “everything that can be measure can be improve “.Mr. Stan said use queuing theory. I found out imformation about this theme and now I have a reseach of 3 Japanese reseachers who apply this theory to reduce waiting time. Mr. Stan thank you.To Seider, thanks so much too, this is what I was a little bit confuse…..this mean you take data according to the combination you have in the DOE.
    Mr. Robert thanks so much for your commends.
    I am not trying to use DOE by force, I just wanted to know perfectly about this theme and how to use this powerful tools. Thanks to all of you I understand how to use it. I agree with Alastair about “Put the poster in the right direction”…..I will be posting question and please let me know your commend as alwaysJose Ramon

    0
    #168472

    Stan”s sister
    Member

    You have to go for 8 runs (minimum),thento decide the higher and lower data.You are completely wrong

    0
    #168489

    Chris Seider
    Participant

    You’re welcome.
    Sincere, clearly stated questions tend to get replies and I attempt to add my part when time and conditions and interest coincide.

    0
Viewing 8 posts - 1 through 8 (of 8 total)

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