In this video, we take a look at transforming data using the QuikSigma software to get effective results within your Six Sigma projects. Get your copy of the software here.
Fairly often, we’ll deal with data that are not normally distributed. A very common example is waiting times, in lines at banks, or to buy tickets at the movie theater, or something. Let’s take a look at our data and see what non-normal data looks like. What we can look at here, is the distribution down here at the end of the behavior chart. This is where we do our capability study.
Interestingly enough, non normality is not a big issue for the behavior chart. It’s only when we get over here and get estimating defective parts-per-million, based on the normal distribution, that things get a little unreasonable, if we’re not normally distributed. For that purpose, QuikSigma provides this transform function. If you look at at this presentation of the data, that’s distinctly very non normal, and all I have to do to normalize it is quick transform. That will raise all of my data, and my control chart limits, and any specifications that I’d entered to the power, that is shown in the lambda box. That is the power that QuikSigma estimates is most likely to give us a normal distribution. Sure enough, it does.
So, if I enter a specification here that will also be transformed, just a minute. One thing we didn’t do, let me put this back. If you ever wanna just put this back to the way it was, just enter a 1 here, and you’re back to your original distribution. With an upper spec limit out of 5, we’re really showing not terribly bad results. As soon as I transform that, my specification will be transformed 1.36, and look what that does to my defective parts per million. Really about two orders of magnitude change, and this would be taken as the more accurate estimate. One other thing that you can do in case you’re curious, if you enter a 0 here, it will take the logarithm of the data. In this case that doesn’t do too badly either. That gives us a fairly normal distribution.