• Zilgo replied to the topic P-value in ANOVA table in the forum General 18 years, 2 months ago

    What used to be done, before computers, was that you would look up a critical value and compare it to your observed F-statistic.  That critical value represented the cutoff point for the alpha level of significance, usually 0.05.  Well, 0.05 is also the p-value that is associated with a certain F-statistic.  When you see a p-value in an ANOVA ta…[Read more]

  • Zilgo replied to the topic ANOVA in the forum General 18 years, 4 months ago

    ANOVA is typically used when you have a continuous output and discrete inputs.  Minitab says you have to have numeric values so if you are using a categorical variable then you can just assign a number to each category and it will work.

  • Zilgo replied to the topic Standard Deviation in the forum General 18 years, 4 months ago

    Why do we need to use any other method when there is standard deviation?
    As long as it makes sense and can be interpreted, then there is nothing wrong with using another measure.  SD is just very common and applies to all statistical testing.

  • Technically, it is mathematically impossible to convert from Z value to DPU.
    UNLESS, you assume a balanced didtribution (ie: your data is centered between Z and -Z)

  • Zilgo replied to the topic I charts in the forum General 18 years, 4 months ago

    There is a big difference between “in spec” and “in control”.  Being in spec doesn’t say anything about whether or not you are in control because spec limits can be arbitrarily placed.  Being in control doesn’t tell you anything about whether you are in spec because the control limits are calculated independently of the specs.  Besides, being in…[Read more]

  • Control charts do NOT measure the sigma level of an overall process.  They measure whether a process is in statistical control, ie: does the project generally follw a normal distribution.  Since +/- 3 sigma encapsulates 99.73% of the data in a normal distribution, if you process falls within that limit, you have a process that is in statistical c…[Read more]

  • Zilgo replied to the topic Normal Distribution? in the forum General 18 years, 5 months ago

    Not so fast, my friend.  Most of the statistical analysis done in six sigma relies on the assumption that your data set is normal.  Too many times people bring stuff across my desk and say that their results are good but when I check the assumptions of the tests they ran, they ignored the fact that their data is non-normal.  They are about a mi…[Read more]

  • Zilgo replied to the topic Setting thresholds in the forum General 18 years, 5 months ago

    Just because a process is in control doesn’t mean there is anything that is worthwhile to do with it.  It just means you don’t (statistically) have to do anything.  Lots of in-control processes are way out of spec or very far from a target.  I guess my point is there is always something you can do.

  • Zilgo replied to the topic BOX-COX WONT TRANSFORM in the forum General 18 years, 5 months ago

    Just because it’s fun to throw out something extra…
    When you have a large dataset that is not normal, most of the time it won’t belong to any other easily recognizable distribution.  I do like Robert’s suggestion.  That book is a good reference for strange capability studies.

  • No, one is definitely made up of a squared sample size, which I have never seen in any application.  Truth be told, I was mildly curious as to where the formula you posted earlier came from.

  • As always, instead of just trying to remember a formula, it’s best to look at a situation and apply what you know.  sd = sqrt(var) = sqrt(SS/(n-1)).  Standard deviations are not additive, and since each sd was based on a different n, in this case the variances are not really additive either.  So the only thing that is additive is the SS (Sums of…[Read more]

  • The StDev(within) is the standard deviation of the measurements within your subgroups.  I think there are five methods you can choose from to figure out how MINITAB calculates the SD.  You can find those choices under the ‘Estimate’ button in your Capability form, and the formulas for those are in the Help File

  • Zilgo replied to the topic Question in the forum General 18 years, 6 months ago

     When I read that I want to say that it can’t be right, and here’s why.  The area under any normal curve = 1.  If you take a portion of the curve, say between +/- 1 sigma, and create a ratio with another curve, say +/- 1 sigma, you get a ratio of 1 = sigma1/sigma2.  But sigma 1 sigma2 necessarily.  You could easily construct examples where it d…[Read more]

  • Since Six Sigma is a method for continuous improvement, presumably no one is ever shooting straight for a 6.0 sigma level throughout the company (unless of course they were already running at 5.9).  If your initiate Six Sigma and measure a company wide sigma level (a questionable practice anyway) of 3.0, you should be targeting projects that w…[Read more]

  • As Erik said, check what the appropriate sample size would be and then as Scott said, use a proportions test.  ANOVA is not ideal in this case since you have discrete data.  Binomial (only choices are a 1 or 2) to be exact.
    It would be my suggestion to actually use a one proportion test.  You would test the proportion of failures from your sa…[Read more]

  • Not to intrude but it seems like you have artificially created normal data through subgrouping.  Something about the way you subgrouped your data makes the distribution look different.  You need to ask yourself whether subgrouping was necessary in this case.
    Scott, as per your comments on transformations, sometimes data is just from a different d…[Read more]

  • Zilgo replied to the topic Normality Tests in the forum General 18 years, 6 months ago

    P-values are used in all hypothesis tests.  Goodness of fit tests are just a type of hypothesis test that has to do specifically with the impact of an X on a Y, used in model building.  But there are many other hypothesis tests besides goodness of fit tests.  One is a normality tests which tells you whether or not your data violates the as…[Read more]

  • The easiest way to accomplish this task is to perform the 2 Proportion test multiple time.  Since there are only three comparisons to perform (1v2, 1v3, 2v3) this isn’t really very tedious or time consuming, as it would be if you had like eight proportions to compare.  This will at least give you the three confidence intervals you need to c…[Read more]

  • You could run a correlation/covariance analysis to see whether or not your Xs are correlated.  A scatterplot of each X versus another will show that if you need it graphically.
    As far as running a regression analysis on discrete X I am going to assume that you really have multiple level categorical data which means you need to run a logistic…[Read more]

  • Zilgo replied to the topic Process capability in the forum General 18 years, 7 months ago

    Z bench is the min(Z USL, Z LSL).  Since half your distribution is below the LSL, the Z LSL is going to be zero.  Since likely all of your data is below the USL, the Z USL is going to be big, probably around 5.  0 < 5 so Z bench is also 0.

  • Load More