Search All of the Math Forum:
Views expressed in these public forums are not endorsed by
Drexel University or The Math Forum.



Re: Isoplot
Posted:
Jul 12, 2006 9:39 AM


Kenny:
Everything Dorian Shainin does is "by the seat of his pants". When subjected to tests to see how well they work, some work rather well and some are very poor.
I've "deprogrammed" a lot of people away from Shainin methods, especially isoplots. While there's nothing fundamentally wrong with isoplots, they are somewhat "iffy". Basically, isoplots are attempting to estimate "components of variance"... breaking down the total variation (total variance) into its component parts. There are better ways to do this. I did some work to test "how good are isoplots" many years ago, and decided that I'd not teach isoplots in the corporation where I worked. The effort required to teach isoplots and "make them stick" was just not appropriate. So I went another route using software that I wrote and distributed for all who cared to use it.... and of course taught how to use it.
All of Shainin's methods are touted as being appropriate for use on a factory floor. That is... can be taught to just about anyone. In the 1950s (when Shainin began teaching these methods... actually pick them up from another person whose name I cannot recall, but he worked at Gillette) these concepts seemed to be revolutionary. By the mid1960s their value had degraded because alternative methods are better. In particular, his concepts for "finding the red X" are quite "iffy". But that's another story.
KennyB wrote: > Does anybody have information (or can point me in the right direction > to find the info.) regarding the statistical validity of using the > Isoplot technique (developed by Dorian Shainin) to determine the > effectiveness of a measurement system?



