iSixSigma

Analyse Phase Question

Six Sigma – iSixSigma Forums General Forums Methodology Analyse Phase Question

Viewing 6 posts - 1 through 6 (of 6 total)
  • Author
    Posts
  • #54221

    narayana prasad
    Participant

    Hi Experts,

    I am trying to solve an exerciser. It would be great full if I can get some solution. How to test the below samples and check whether a given characteristic of 5 is maintained on the product. The below is the data. And excel file is attached as well.

    Values Mat-A Mat-B PriceIndex Achievement DBP1 DBP2 Growth Treatment EnzymeActivity Therapy Litter Response Impact HelmetType Location
    4.9 13.2 14.0 0 77 90 62 15.1 1 0.15 1 1 0 47 1 1
    5.1 8.2 8.8 50 88 72 85 13.0 1 0.26 1 2 0 43 1 1
    4.6 10.9 11.2 56 85 61 78 16.2 3 0.23 1 3 1 46 1 2
    5.0 14.3 14.2 72 74 66 66 14.9 1 0.99 1 4 0 40 1 2
    5.1 10.7 11.8 80 75 81 80 13.2 1 0.55 2 1 1 62 2 1
    4.7 6.6 6.4 80 62 69 91 13.8 3 0.26 2 2 1 68 2 1
    4.4 9.5 9.8 80 80 59 69 13.1 2 -0.22 2 3 1 67 2 2
    4.7 10.8 11.3 99 70 70 77 13.0 2 0.99 2 4 1 71 2 2
    4.6 8.8 9.3 101 83 84 12.9 2 0.55 3 1 2 41 3 1
    13.3 13.6 110 11.9 1 0.66 3 2 3 39 3 1
    110 17.0 3 0.77 3 3 3 42 3 2
    110 12.8 2 0.99 3 4 2 46 3 2
    120 14.7 3 0
    140 12.0 2 0
    144 15.0 3 0
    145 16.5 3 2
    150 1
    180 1
    201 1
    210 1
    220 2
    240 3
    290 3
    309 0
    320 0
    325 1
    400 0
    500 2
    507 2
    3

    0
    #194281

    Robert Butler
    Participant

    You will have to provide a lot more detail and background before anyone can offer much of anything in the way of advice. In particular the relationships between the various columns and how they might relate to “a given characteristic of 5” (is this “characteristic” supposed to be the column headed “Values” or is this something else?)

    Your excel data set looks like you only have nine (samples?) with a complete listing of parameters(?),variables(?), attributes(?)and one that has everything except an entry for DBP1. Ten of the samples(?) have a number for the column headed “Values”. Is this correct or is it something else?

    0
    #194283

    narayana prasad
    Participant

    Hi Robert,

    Thank you very much for your reply. That is correct the characteristic is the column headed “VALUES”. I have only got this info and have to solve this as part of my exercise. Assuming column ‘vales’ as our desired characteristic can we test these samples to find 5 is maintained across the product.

    Thanks again for you help.
    Regards,
    Prasad

    0
    #194284

    Robert Butler
    Participant

    You still need to explain what you mean by “can we test these samples to find 5 is maintained across the product.”

    1. Are you looking for deviations from some average?
    2. Are you looking for changes in the numbers as a function of some of the other columns?
    3. Are you looking at time trending (here we’re assuming the data entries are in time order which may or may not be true)?
    4. Are you supposed to try to build some kind of predictive model using some of the other parameters to see if there are combinations of these variables that will always result in a “5”?
    5. If it is a predictive model what have you been told with respect to variable choice? You have more variables than recorded runs so if this is the issue you are going to have to triage those variables.

    As I said in the first post – you need to provide a lot more detail before anyone can attempt to answer your question.

    0
    #194286

    Chris Seider
    Participant

    @rbutler You are very patient.

    @nkanchar Consider attaching your project charter including your problem statement.

    0
    #194289

    Robert Butler
    Participant

    So, basically it looks like you have been given a bunch of “stuff” and it’s your job to do the equivalent of an English III class (creative writing). So the most obvious eyeballing of the data using the one-factor-at-a-time-while-pretending-nothing-else-matters-and-it-isn’t-confounded-with-everything-and-its-grandmother approach is to insist on using only Treatment value of 1 and ignore Treatment value of 2 and 3.

    I’m guessing that what they want you to do is run a bunch of one-factor-at-a-time regressions of all of the stuff against values and see which one provides a significant correlation. Just giving the data a visual lookover I think the only one that will have a p < .05 will be Treatment and from there the statement above will follow. I'd recommend trying this approach (checking for significance of univariate regressions) and see what you see.

    Just FYI the recommendation of running simple linear regressions is assuming the numeric values for things like Treatment are actually measures of some kind and not simple identifiers of type. If the values for things like Treatment, Therapy, Helmet Type, and Location are nominal catagorical variables then you would need to build dummy variables for any of these with more than two levels and run the regression against the dummy variables.

    Given the nature of the posts you have made to this forum my guess is that you haven't had an introduction to dummy variables and their use so, again, I'm guessing, your instructors probably won't mark you down for running the analysis as a series of simple univariate regressions.

    If you're interested in learning more about dummy variables (you should be) I'd recommend reading Regression Analysis by Example by Chatterjee and Price – they have a good section on this issue.

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

You must be logged in to reply to this topic.