Philosophy of Experimental Design — Transcript

Full transcript

1. 0:02hello my name is Tom Berker I'm the
2. 0:05author of the book quality by
3. 0:06experimental design which is in its
4. 0:08Third Edition I retired from teaching
5. 0:11after 35 years at the Rochester
6. 0:13Institute of Technology and have the
7. 0:15title professor emeritus before RIT I
8. 0:20spent 21 years at Xerox as an engineer
9. 0:23and utilized statistical experimental
10. 0:25design extensively in over 2,000
11. 0:28projects during the early de development
12. 0:30of the zerog graphic process what I want
13. 0:32to do with these introductory viets on
14. 0:35statistical experimental design is to
15. 0:37introduce you to the power of these
16. 0:39methods that link the three essential
17. 0:42elements of bringing a product to the
18. 0:45marketplace Engineering Management and
19. 0:49statistics by managing the engineering
20. 0:51activity with the power of statistical
21. 0:54thinking not only do you have the
22. 0:56ability to create a system to
23. 0:59commercialize the product you do so in
24. 1:02the most coste effective manner possible
25. 1:05and guarantee the quality of that
26. 1:07product all at the same time let's begin
27. 1:12by looking at the philosophical reasons
28. 1:14behind using statistical experimental
29. 1:17design I've brought my monk Friends
30. 1:19Along since they are such great
31. 1:21philosophers my monk friend asks why
32. 1:24design experiments I'm sure you have
33. 1:26your own personal reasons but here are
34. 1:29four reasons I I have found to be both
35. 1:31Universal and extremely important first
36. 1:34we design experiments because this
37. 1:36method gives us a structured plan of
38. 1:38attack on the opportunity before us
39. 1:40second because these are statistical
40. 1:43experimental designs we can easily mesh
41. 1:46proven statistical analysis methods to
42. 1:49determine the most likely outcomes third
43. 1:52statistical experimental designs are
44. 1:54inherently more efficient now I'll go
45. 1:57into excruciating detail on this point
46. 1:59so since it is Central to the management
47. 2:02and success of experiments and fourth
48. 2:04because we have a structured plan of
49. 2:06attack we are forced to get organized in
50. 2:09our
51. 2:10experimentation the greatest reason for
52. 2:12failure in experimentation is lack of
53. 2:15organization there is an entire vignet
54. 2:18devoted to this topic where you will
55. 2:21learn how to
56. 2:23organize here is the definition of
57. 2:25efficiency an efficient experiment gets
58. 2:28the required information at the least
59. 2:30expenditure of resources this is an
60. 2:33exact definition and has three elements
61. 2:36of an efficient experiment the first is
62. 2:39experiment there's a big difference
63. 2:41between a test and an experiment second
64. 2:44we get the required information not too
65. 2:46much not too little just right and
66. 2:49finally we get this information at the
67. 2:51least resources the key elements then
68. 2:54are
69. 2:55experiment required and resources
70. 3:00let's look at the difference between a
71. 3:02test and an experiment testing is an
72. 3:05Endeavor to see if something works now
73. 3:08my mon friend has devised a Dead Sea
74. 3:10scroll translation device that does not
75. 3:13seem to be working why of course it's
76. 3:16not plugged
77. 3:18in so a monk who is a member of the
78. 3:20electrical union comes and connects it
79. 3:23to the mains the knife switches thrown
80. 3:28and poof the whole thing blows up have
81. 3:32you ever experienced a poof test the
82. 3:35problem is there is usually nothing left
83. 3:37to analyze why it poofed experimentation
84. 3:40on the other hand links the ifs of
85. 3:43testing to determine why why is the most
86. 3:47important question you can ask as you
87. 3:50will see an experiment is a structured
88. 3:53set of coherent tests that are analyzed
89. 3:56as a whole to gain understanding of the
90. 4:00process this is how experimentation
91. 4:02differs from testing because
92. 4:04experimentation is oriented to
93. 4:08understanding understanding breathes
94. 4:10control and leads to Lasting success a
95. 4:14successful test may give us an instant
96. 4:16gratification today but may lead to
97. 4:19unfortunate surprises tomorrow an
98. 4:22experiment gives us a functional
99. 4:24relationship an equation that leads us
100. 4:27to product and process design designs
101. 4:30that have no
102. 4:32surprises oh it looks like my monk
103. 4:34friend used an experiment his
104. 4:37translation machine is
105. 4:39working if I have convinced you that you
106. 4:42should be experimenting not merely
107. 4:44testing then let me show you the three
108. 4:46elements of a good experiment you need
109. 4:49knowledge of the process clear goals and
110. 4:51objectives and a response variable but
111. 4:54you say but I'm experimenting to gain
112. 4:57knowledge experimental design focuses
113. 4:59the the prior knowledge you already have
114. 5:01the things you learned in school
115. 5:03informal on job training even small
116. 5:06preliminary experiments or even tests
117. 5:09then there is the lore of accumulated
118. 5:11data this can be used as a numerical
119. 5:13brainstorming to uncover possible
120. 5:15factors that need to be studied in a
121. 5:18structured manner unfortunately I have
122. 5:20seen statisticians waste their time
123. 5:22trying to make sense out of messy plant
124. 5:25data use plant data to uncover likely
125. 5:28factors more than that is usually
126. 5:33impossible it is important to separate
127. 5:35the goal the end result from the
128. 5:37objective how we get there in fact the
129. 5:39objective defines the required
130. 5:41information here's an example of a goal
131. 5:44and objective taken from the
132. 5:46photographic chemistry industry the
133. 5:48picture on the left is washed out and
134. 5:50not very pleasing our goal is to make
135. 5:52the image quality better like the
136. 5:54picture on the right the objective
137. 5:57however defines what we are going to
138. 5:59look at in the experiment we have come
139. 6:01up with a set of factors that our prior
140. 6:03knowledge says will influence image
141. 6:05quality that's IQ we want to test the
142. 6:08idea hypothesis is a big word for idea
143. 6:12that IQ is functionally related to the
144. 6:15amount of developer Elon and
145. 6:20hydrocone the amount of preservative
146. 6:23sodium sulfide the amount of alkali to
147. 6:26make the developers work and the amount
148. 6:28of inhibitor
149. 6:30plus the temperature of the solution and
150. 6:32the speed at which the photographic
151. 6:34material travels through this solution
152. 6:37notice the prior knowledge that goes
153. 6:39into assembling these
154. 6:42factors for a successful experiment we
155. 6:45need something to measure a response
156. 6:47variable or possibly
157. 6:50variables such responses need to be
158. 6:53quantitive precise and
159. 6:56meaningful a quantitative response is so
160. 6:59much better than good better best such
161. 7:02words are very uncertain numbers are
162. 7:05easily understood across continents and
163. 7:07countries and of course we need the
164. 7:09numbers to compute the statistics when
165. 7:11we analyze our data be sure to
166. 7:14standardize these responses and avoid
167. 7:16engineering Wars that debate how to
168. 7:18measure and waste so much time and
169. 7:22effort a precise repeatable response is
170. 7:25essential accuracy or how close the
171. 7:28response is to to the true value is
172. 7:30important but not essential here's the
173. 7:33difference between accuracy and
174. 7:35precision to illustrate let's go to an
175. 7:37archery match in Sherwood Forest our
176. 7:40first shooter is frier tuck his arrows
177. 7:44are all over the place never hitting the
178. 7:46target he has neither accuracy nor
179. 7:50Precision Ellen adale is next and his
180. 7:53sights are slightly misaligned he just
181. 7:55got a new bow his cluster is good but he
182. 7:58is not exactly on the
183. 8:00bullseye of course Robin Hood strikes
184. 8:04the center of the target over and over
185. 8:06and over again splitting his arrows
186. 8:08Robin has both accuracy and precision
187. 8:10and a large bill for
188. 8:12arrows if of course we could aspire to
189. 8:16Robin Hood's accuracy and precision that
190. 8:18would be great for our response but if
191. 8:21we can live up to alen ad Dale's
192. 8:22Precision will do okay in experimental
193. 8:27design the final quality of a good
194. 8:29response is that the response is related
195. 8:32to the customer
196. 8:33requirements who is the
197. 8:36customer this is not necessarily the end
198. 8:40use customer but the customer who will
199. 8:42get the output of the experiment we are
200. 8:44working on the next in line to assure
201. 8:49this happens the customer of the
202. 8:51experiment must be a part of the
203. 8:52experiment and make known the responses
204. 8:56needed this is part of the organization
205. 8:59of EXP
206. 9:02experimentation let's look at the
207. 9:03definition of an efficient experiment
208. 9:05once again and focus on the final
209. 9:08Element Resources what are
210. 9:12resources money is certainly a resource
211. 9:15it is a renewable resource we can borrow
212. 9:18money or invest it from profits people
213. 9:21are a resource we can hire new people or
214. 9:24redirect people from other projects
215. 9:26another renewable resource but time is
216. 9:30something we can neither beg borrow nor
217. 9:32steal time is the most precious of
218. 9:35resources statistical experimental
219. 9:38design conserves all of these resources
220. 9:43especially
221. 9:47time