Laserfiche WebLink
<br />B-9 <br /> <br />Post hoc analysl~s were required to place the experimental days in meteorological seedabil- <br />ity categories consistent with those specified for the experiment. This was at least in part due <br />to the difficulties of forecasting cloud top height over a 24-hour period. <br /> <br />* Israeli II * <br /> <br />With only limited detail available, no purely statistical questions have been identified. <br /> <br />* Alberta hail * <br /> <br />No randomizatiion, hence, no firm statistical conclusions. <br /> <br />* Santa Barbara II * <br /> <br />The analysis of phase 1 has been attempted in various ways, of which two are discussed in <br />more detail in Section 42, in the appendix. Significance hovers near the conventional 5% point. <br /> <br />Phase 2 was terminated before it was reasonable to anticipate statistical signific:ance. <br /> <br />If the raingauge-by-raingauge analyses of Phase 1, ending in contour maps for both <br />apparent ratios and indicated significance, are to be adequately evaluated, we believe it will be <br />necessary to conduct re-randomizations and draw the resulting contour maps. Initially, it would <br />be very helpful to do this for as few as 25 re-randomiza.tions (or perhaps fewer), with the anti~ <br />cipation that examining these results would guide the decision as to how many mon~ were <br />worthwhile. <br /> <br />* South Africa hail * <br /> <br />No randomizatiion, hence no firm statistical conclusions. <br /> <br />* North Dakota * <br /> <br />There was no clear choice of a main criterioI1l in advance. However, similarity of <br />p-values for the three main criteria suggests that correlation is high enough to make multipli- <br />Clty from this :source not a serious problem. Results of analyses for all days were far from <br />significance. <br /> <br />Results for separations were sometimes individually significant, but the numbers of alter- <br />natives considered were large enough to leave no trace of simultaneous significance. <br /> <br />We conclude that there was no evidence for confirmation though some separations were <br />suggestive and perhaps encouraging. <br /> <br />* Tasmania * <br /> <br />The most sharply-focussed and firmly-attacked question available from the publishf:d ana- <br />lyses involves (their Appendix 2, Table 4): <br />1) the whole targe:t <br />2) either untransformed or cube-root transformation <br />3) the regression analysis <br />4) the permutatiolll test <br />5) the four seasons <br />Here it seems reasonable to average over the two transformations and regard the seasons as <br />n~cessary multiplicity. The 1Ulting tail areas are Autumn, ~ (.04+.11) = .075; Winter, <br />~ (.18+.09) = .135; Spring, 2(.98+.80) = .89; Summer, ~ (.41 +.44) = .425. lj. we regard <br />these as order statistics of a sample 01"4 from the uniform distribution, we find (see Section 38 <br /> <br />1\ <br />