Downloaded from https://www.everyspec.com AMCP 700.198 Example No. 15 (Conttd) 9. W•ite the cquation for 9. Var{y*(x•}lw 1672.5 + 0.2247x. + 50175.5 Var {y*(.)}. -51, 848 + 0.2247x'. 10. Calculate Var {y*(x)) 10. x - 200 for several x. Var {y*(200)} - 51848 + 8988 - 60886 2471 Var {y*(100)1 - 54095 - 2332 Var {y*(0)} - 2283. 11. Calculate the linear 11. Zt(Yj - b)1 - 31, 877, 000 correlation coeffic lent. Use Eq. 2-66c. e- b)Z/(- - 2) ='31, 877, 000/28 6 (1066.99) i - 2 0 0o6.)".956 "a - 0.978. 12, Use Table 15 of Ref. 30 12. The chart is not clear for such a high but it to get 95% a-confidence appears that Conf {0.95 S p S 0.99} R 95%. In limits for p.
any event, the linear correlation is very high. Virtually all tho original variance in y is ex- plained by the regression. The statistical data have all been gathered. It now remains to interpret it. With Y - 28, it makes little difference whether Student "-distribution is used, or the #-nor- mal distribution. One could use a goodness-of-fit test on the e; to see 1f they could reasonably have come from a s-normal distribution. More disturbing are the very high deviations of blade 6, and the fact that the I c I appears greater for the half of the sample with higher resonant frequency.
If the engine operating requirements can be satisfied with the present knowledge, no further experimental or statistical tests are needed. But remember, 1/30 of the sample had a deviation exceeding k 3 standard deviations. Do not blindly forget the sample and then make predictions of the population based on some abstract statistical procedures. Statistics can only answer the questions it is asked. If you don't ask the "irgh't" questions, you will get answers to irrelevant questions. X I ....... . .> ~ ; .......... ..... ........ ......