AL
saying bioassayists are untrained, inept, or sloppy—on the contrary! I have great admira- tion for their heroic efforts, given what they have to work with. It’s really tough to be a precise bioassayist!
But getting back to Figure 4, it’s my judgment that many biopharmaceutical manufacturing processes are probably inherently less variable than they appear. Biopharmaceutical engineers don’t get to see what good engineers they are, because the measurement variation masks the true tight behavior of their fine industrial processes.
in this picture: σT = √(12 so Cpk = (L – U)/(3 × σT
My second point has to do with the apparent Cpk
+ 22 ) = 5/(3 × √5) = 0.75 <
1, which suggests that the industrial process is not capable. In fact, the true Cpk
of the in- dustrial process is 5/(3 × σP) = 5/(3 × 1.00) =
1.67 > 1, which means the industrial process is inherently highly capable—it only appears to be incapable because of the obscuring measurement process variation.
I once read a talk from an awards ceremony for a statistician, during which he basically said to his colleagues, “Hey, guys. I just discovered that if you clean up your analytical methods, you can improve the Cpk
of your industrial pro-
cess without ever touching the process itself!” Well … yeah.
When I started graduate school at Purdue in 1966, we incoming analytical chemistry students were given a pep talk by Professor Herbert W. Laitinen,1
who was visiting from the
University of Illinois. For some reason, I always remembered one of the things he said during that inspiring talk: “When you develop an analytical method, it should have only about one-tenth the variation you expect from the industrial process for which it will be used.” I never fully appreciated that statement until I started to look at process capability. If σP and σM = 0.1, then σT = √(12
= 1 + 0.12 ) = √1.01 =
1.005. The measurement process now has very little effect on the total variation!
This is illustrated in the simulation shown in Figure 5. The measurement process variation is one-tenth the industrial process variation, and the observed behavior of the industrial process (the green squiggly line at the very
top right in the figure) matches almost exactly the true behavior of the industrial process (the red squiggly line in the top graph above the industrial process). Practically speaking, this is what the engineers need and want, the thought we’ve been holding from our discussion of Figure 2.
Fifty years later, it’s good to appreciate that the pioneers in the field knew exactly what they were talking about. We should keep these con- cepts alive.2,3
References
1. Laitinen, H.A. Chemical Analysis: An Advanced Text and Reference; McGraw-Hill: New York, N.Y., 1960.
2. Hayes, J.D. Is it Process Variability or Mea- surement System Variability? http://virtual.
auburnworks.org/profiles/blogs/is-it-pro- cess-variability-or-measurement-system- variability
3. Rodebaugh, B. Six Sigma in Measurement Systems: Evaluating the Hidden Factory,” slide 7;
http://asq.org/cpi/2002/06/six- sigma-in-measurement-systems-evaluat- ing-the-hidden-factory.ppt
Dr. Stanley N. Deming is an analytical chemist mas- querading as a statistician at Statistical Designs, 8423 Garden Parks Drive, Houston, TX 77075, U.S.A.; e-mail:
statisticaldesigns@gmail.com; www.
statisticaldesigns.com
AMERICAN LABORATORY 25 JANUARY/FEBRUARY 2017 Figure 5 – The world according to Laitinen. ) = √5 = 2.24,
Providing a Macro View on Micro Analysis
Proven,
Robust, Reliable Thermal Analysis Solutions
SHIMADZU AIMs to provide analysis systems for all users. Used in conjunction with the IRTracer-100 QT +4#HƂ PKV[
5 (6+4
spectrometers, the fully automated AIM-9000 infrared microscope allows performing all steps in micro analysis quickly and easily.
X Observe X Measure X Analyze
Simplify micro analysis with:
• 9KFG (KGNF %COGTC YKVJ : \QQO HGCVWTG – makes for fast and easy positioning for accurate analysis
• #WVQOCVKE %QPVCOKPCVKQP (GCVWTG – eliminates manual search and masking routines for faster analysis
• 2TQRTKGVCT[ %QPVCOKPCPV #PCN[UKU 2TQITCO – analyzes contaminants with a proprietary search routine, providing fast, accurate search results 5 0 TCVKQ – yields low noise
•
spectra for even the smallest contaminants for better library matching results
Learn more. Call (800) 477-1227 or visit
www.ssi.shimadzu.com/AIM
5JKOCF\W 5EKGPVKƂ E +PUVTWOGPVU 4KXGTYQQF &T %QNWODKC /&
+PHTCTGF /KETQUEQRG
#+/
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56 |
Page 57 |
Page 58 |
Page 59 |
Page 60 |
Page 61 |
Page 62 |
Page 63 |
Page 64 |
Page 65 |
Page 66 |
Page 67 |
Page 68