• Evil replied to the topic Communication Questions in the forum General 14 years ago


  • Evil replied to the topic Basic Question in the forum General 14 years, 3 months ago

    there’s so many ways to do this and all depend on the circumstances at hand so I’ll give you my opinion.
    In the beginning of a project when I’m trying to establish a baseline, if I can collect the x’s that are easy to get and likely candidates along with the y, then I’m going to do so. If I get lucky and find some strong correlation,…[Read more]

  • A few questions…with the possible outputs being between 0 and 8, did your observed outputs end up utilizing the full range ?
    A few things to try…
    1. Ordinal Logistic Regression – your output basically sounds ordinal
    2. Rank transformation on the data then re-analyze using GLM the transformed y.
    Missing an x is always a possibility but you…[Read more]

  • Evil replied to the topic Invoice Accuracy – DPMO in the forum General 14 years, 4 months ago

    It depends (the usual answer)… If you track defective invoices (3/100 bad), you’ll see the quality from the customers view assuming that an invoice with 1 or more defects is unacceptable from their standpoint. If you track defects (10/1300), then you’ll have a much more sensitive measure of whether you’re improving. Have a look at the…[Read more]

  • Evil replied to the topic Triz and SS in the forum General 14 years, 5 months ago

    Yes…lots of experience doing applying TRIZ in SS/Lean projects in all industries…
    In Define: use IFR and 9 Screens(are you solving the right problem ?)
    In Improve: use all the TRIZ tools to come up with innovative ways to solve the problem now that you know the causal relationship (y=f(x)).If you haven’t had much success, then your original…[Read more]

  • Evil replied to the topic Process Capability in the forum General 14 years, 5 months ago

    Jessd,You’ll also need the number of total bills processed to know the opportunity size. e

  • Evil replied to the topic DOE in the forum General 14 years, 5 months ago

    If the 10 responses you recorded were taken without ever changing any of the factors, then these are repeats. (thus, no extra DOF). With repeats, you can take the average and take the natural log of the standard deviation. The first will be a model that contains factors shifting the mean, the second a model containing factors that affect the…[Read more]

  • Face to face is always going to be better for so many reasons (see the process, see their data, show how you are there to help them, etc). Remote works as long as you’ve developed a relationship and established some credibility. Frequency is important (at least weekly) whether remote or f2f.
    I make use of on-line meeting tools like gotomeeting…[Read more]

  • Nelson,
    There are quite a few online training programs that will give you a so called “certification” after passing a test. You have to ask yourself why you need certification. If it’s just a credential needed to advance in your organization, find the least expensive one and do it. If you want to really learn and become a strong GB/BB/MBB,…[Read more]

  • Evil replied to the topic Process Capability in the forum General 14 years, 12 months ago

    When calculating process capability and the data appears to be non-normal, the first order of business is to ask yourself if the data should be inherently non-normal. Sometimes, the data appears non-normal only because its so unstable OR a mix of different process conditions. In the case of cycle time data on a business process,…[Read more]

  • Evil replied to the topic Is it DFSS? in the forum General 15 years ago

    Yep, this is always the tough part. Sometimes, you can create an estimate from similar processes. Alternatively, you can mock up the process and collect some data.

  • Evil replied to the topic Is it DFSS? in the forum General 15 years ago

    Initially, data from your benchmark process will allow you to estimate(flow up) capability on your future process. The purpose being to adjust the design if the capability is not what you desire. Once you have predicted capability that is acceptable, then you pilot the new design to validate your prediction.

  • Evil replied to the topic Adjusted R-squared in ANOVA in the forum General 15 years ago

    Hi Nate,R-squared is the percentage of variation in your data that your “factor” serves to explain. In this case, your grouping factor in the 1-way anova basically explains none of the overall variation.When doing any hypothesis test, if you didn’t “engineer” the sample size to give you a desired amount of power, then failing to reject the null…[Read more]

  • Evil replied to the topic Is it DFSS? in the forum General 15 years ago

    Flowing up capability means to “predict” your process capability BEFORE actually implementing the new process. In order to do this, you obviously need data on the sub-elements of your process so you can do some type of simulation (i.e. Monte Carlo).

  • Evil replied to the topic Is it DFSS? in the forum General 15 years ago

    Technically, no…however, in taking an existing process and adapting it to your division, you can utilize some of the DFSS tools. If you collected some VOC at your division, distilled that into process requirements, designed the process to meet those rqmts at a high level of capability, (even using elements of the similar processes), flow up…[Read more]

  • Lisa,
    If you have overall satisfaction, you can use Ordinal logistic regression to see what other question responses tend to correlate best with the OA Sat.A piecemeal approach is to use Chi square test of association, basically analogous to doing each x vs the y.Hope that helps…

  • Evil replied to the topic DOE in the forum General 15 years, 2 months ago

    Ravi,It depends…it depends on so many things that it would be tough to say whether it’s OK or not. That’ll be a res V design so it’s a pretty safe fractional design (mains and two-ways reasonably clear).You might do some additional research on DOE or seek someone that can help mentor you though this expmt.E

  • I would suggest estimating what the distribution of runout is (could be any number of zero bounded distributions, if you had some data from the process that creates it, that’d be best), then utilize a Monte-Carlo analysis to come up with the resulting distribution of the dimension(gap ?) you’re interested in.
    Hope that helps…

  • If you can actually measure the time (in appropriate units) for each shipment relative to the promise date, you’ll have a much better estimate of process capability. For example, if a particular shipment was 2 days early, t = 2 days. If it was 3 days late, t = -3 days.
    If you turn your data into pass/fail, you’ll have to gather much more data…[Read more]

  • Evil replied to the topic Box-Cox Transformation in the forum General 15 years, 2 months ago

    Exactly right…when I encounter BB’s that have non-normal data, the first question is’ “Would you expect the process to produce that ?”, or “is that data non-normal because of special cause variation ?”.
    If truly the process produces non-normal data and you’re trying to calculate process capability, then great, use Box-Cox, or one of the other…[Read more]

  • Load More