iSixSigma

Hypothesis Testing

Viewing 5 posts - 1 through 5 (of 5 total)
  • Author
    Posts
  • #55060

    Prabhu V
    Participant

    Hello All,

    I would like to hear from the forum members’ viewpoint on applicability of Chi-square test on the following situation.

    On one of my study/assignment, I need to prove that hypothesis (Assumption) is correct or not (Binary or categorical)

    like Product recalls (taking back the product to company from market) would influence the brand image/sales/market share.

    Since for above, it is very easy to use the Chi-square test to conclude the hypothesis.

    However, I am looking for forum members’ view on the same and as well as suggestions/alternatives for the above situation.

    Thanks in advance.

    Regards

    0
    #198442

    Robert Butler
    Participant

    As you have described it there isn’t much to say except that your view of what you think you are testing/describing is in error. If all you have is product recall yes/no and a list of brands for which you have recall yes/no counts the only thing a chi-square test on a 2 x N table of this type is going to show is a test of significance (or lack thereof) of an ASSOCIATION between recall yes/no and brand type – cause and effect – i.e. recall counts impacting choice of brand cannot be assessed with this kind of analysis.

    0
    #198457

    Amit Kumar Ojha
    Participant

    Hi Prabhu,

    Another line of thought could be to consider the percentage of products recalled and finding the relationship with Market Share for each brand. Here considering both as continuous data, you can use correlation/regression to find the relationship.

    Hope it helps !!!

    0
    #198462

    JB
    Participant

    There will be a lot of factors impacting this analysis. I wonder if it is possible to do a paired comparison? If you could get data on customer satisfaction and/or perception of product quality on the same products directly before and then again after a recall was issued? If possible, that might be the most direct way to see the influence of the recall.

    0
    #198467

    Norbert Feher
    Participant

    Hello,

    Do you have numeric specification criteria regarding to the failures?
    I would make a pareto by brand, product family or other attributive characteristic and go for a more detailed analysis in case if there is no independence.
    To me a Chi square test is also a “quick and dirty” tool to narrow down my search before going into the details…
    I encourage my GB students to use Chi square tests extensively even in excel because count data are almost always available and it is quite easy to understand…

    Regards,

    Norbert

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

You must be logged in to reply to this topic.