iSixSigma

Discrete Y, Continuous X – BLR

Six Sigma – iSixSigma Forums General Forums Methodology Discrete Y, Continuous X – BLR

Viewing 5 posts - 1 through 5 (of 5 total)
  • Author
    Posts
  • #53777

    Heeb
    Participant

    I’m trying to understand if employee experience has an impact in their likelihood to produce defective widgets.

    I have, on a per-employee basis:

    Widgets
    Defective widgets
    Lifetime widgets
    Lifetime widget hours

    What kind of tests can I do?

    Thanks!

    0
    #191418

    HopeOverExperience
    Participant

    I would then look at this to see if there is a difference in the proportions between employees. This will tell you if there’s anything to look at. The test you need will vary depending on the number of employees but probably Chi Square.

    If there is a statistical difference between the people I would look at a regression of % defects against lifetime widgets or lifetime widget hours.

    That should start to give you an indication if there is a relationship.

    0
    #191419

    Robert Butler
    Participant

    I think you will need to provide more information before anyone can offer much of anything in the way of suggestions.

    You said

    “I have, on a per-employee basis:

    Widgets
    Defective widgets
    Lifetime widgets
    Lifetime widget hours”

    What do you mean by the following:

    Lifetime widgets?
    Lifetime widget hours?

    The way it is written it sounds like you have a variety of widgets (Widgets). You have the number of widgets made by a particular employee (Lifetime widgets?) and you have a count of the number of defective widgets made by that employee in their lifetime (Defective widgets) and then, as a guess the total number of hours a given employee has spent on each class of widgets (Lifetime widget hours?). Is this the case or do you mean something else?

    0
    #191420

    Heeb
    Participant

    Hi Robert,

    The Lifetime data are measures of employee experience: The total number of widgets they have produced, and the total number of hours they have spent working on those widgets. Fairly autocorrelated, but we have a decent spread between operators.

    0
    #191423

    Robert Butler
    Participant

    I’m sorry to keep asking questions but I’m still getting two different impressions of the kind of data you have. In your last post the first part of your answer gives the impression that the lifetime data is longitudinal – that is you have counts of widgets made/defects occurring at various time intervals across the lifetime of an operator. But the second part gives the impression that all you have is a single number per widget type: ratio of total defects/total parts and the total amount of time an operator worked on that particular widget type.

    If the data is the former then you have repeated measures data and, if you have the correct software you could build a model to test the effect of time within or even across widgets to see how that correlated with number of widgets produced and number of defects incurred or you could split the “senior” time into a “beginning part of the learning curve” and a “seasoned part of the learning curve” and compare new hire/new widget maker performance against the senior beginning part of the learning curve as well as senior beginning to senior seasoned.

    If the data is the latter then about all you could do would be to look at ratios as a function of time.

    The problem in both cases is you will be confounding all of the changes in production methods, raw material lot changes, etc. with operator length of time since different operators would have started at different times in the widget production cycle and given the state of the production line this could either benefit or hinder with respect to defect counts for the measured periods.

    Either way, if you can provide additional clarification perhaps I or someone else could offer additional thoughts.

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

You must be logged in to reply to this topic.