A little information will allow us to do even better Tempera ture and voltage

A little information will allow us to do even better

This preview shows page 46 - 48 out of 68 pages.

most problems. A little information will allow us to do even better.) Tempera- ture and voltage in the reliability testing of electronic equipment is another. As in so many of the instances we have described above, the main need in this field is for techniques of indication, for ways to allow the data to express their apparent character. The need for significance and confidence procedures will only begin to arise as respectable indication procedures come into steady use. 34. Sizes, nomination, budgeting. The choice of qualitative and quantitative aspects of coordinates is not the only way in which to approximately exercise judgment in approaching the analysis of multiple response data. The work of Dempster (1958, 1960) and of Wilk and Gnanadesikan (1962) points the way toward what seems likely to prove an extensive use of judgment-selected measures of "size" for differences of multiple response. The considerations which should be involved in such choices have not yet been carefully identified, discussed, This content downloaded from 189.63.131.205 on Fri, 29 Jul 2016 14:06:03 UTC All use subject to
Image of page 46
46 JOHN W. TUKEY and compared. Still, it is, I believe, clear that one should not limit oneself to information and judgment about the actual variability, individual and joint, of the several responses (or of more useful coordinates introduced to describe these responses). It will also be wise and proper to give attention to what sorts (= what directions) of real effects seem more likely to occur, and to what sorts of effects, if real, it is more likely to be important to detect or assess. The problems which arise in trying to guide the wise choice of "size" are new, but not wholly isolated. The practice at East Malling, where experiments take a major fraction of a scientific lifetime, of nominating (cp., Pearce, 1953) oertain comparisons, appears to have been the first step toward what seems to be an inevitable end, the budgeting of error rates in complex experiments. We consider it appropriate to combine subject-matter wisdom with statistical knowledge in planning what factors shall enter a complex experiment, at how many versions each shall appear, and which these versions shall be. This granted, how can there be objection to using this same combination of wisdom and knowledge to deter- mine, in advance of the data, what level of significance shall be used at each of the lines of the initial analysis of variance. If wisdom and knowledge suffice to determine whether or not a line is to appear in the initial analysis, surely they suffice to determine whether 5 %, 1 %, or 0.1 % is to be the basis for immediate attention. Yet budgeting of error rate does not seem to have yet been done to any substantial extent. 35. A caveat about indications. It may be that the central problem of complex experimentation may come to be recognized as a psychological one, as the prob- lem of becoming used to a separation between indication and conclusion. The physical sciences are used to "praying over" their data, examining the same data from a variety of points of view. This process has been very rewarding, and
Image of page 47
Image of page 48

You've reached the end of your free preview.

Want to read all 68 pages?

  • Summer '18
  • mit
  • Statistics, E. S. Pearson

  • Left Quote Icon

    Student Picture

  • Left Quote Icon

    Student Picture

  • Left Quote Icon

    Student Picture

Stuck? We have tutors online 24/7 who can help you get unstuck.
A+ icon
Ask Expert Tutors You can ask You can ask You can ask (will expire )
Answers in as fast as 15 minutes