Laserfiche WebLink
<br />A-5 <br /> <br />involved is the same: (j) separating the days (or other experimental units) according to charac- <br />teristics, such as temperature at a specified location in the. cloud, that might reasonably -- so far <br />as our knowledge of the physical phenomena is concerned -- influence the efficacy of seeding <br />and then (ij) examining the apparent efficacy of seeding in each of the separated parts. This <br />will be repeated for a variety, small or large, of different separations. <br /> <br />If no separation that seems even semi-sensible offers a portion in which s(~eding shows a <br />reasonable apparent efficacy, the picture is dark indeed. <br /> <br />If separation unrestrained by physical plausibility produces a separation with a very attrac- <br />tive appearance, the statistician ought to urge extreme caution, recognizing that trying enough <br />randomly-chosen separations will ENSURE finding at least one that appears highly attractive. <br /> <br />In judging the probable repeatability (as in a new confirmatory phase) of an atltractive <br />separation, great weight has to be laid <br /> <br />1) on the physical reasonableness (better, of course, the physical convincingness) of that <br />separation as a way to influence efficacy of seeding. <br /> <br />2) on the extent of supportive evidence available from the after-the-fact analysis of experi- <br />ments in closely related circumstances, if any such are available. <br /> <br />Evaluating how encouraged we ought to be about the promise of a separation is thus difficult <br />and, to a large degree, not just a statistical problem. <br /> <br />It is natural to hope for more than encouragement from separation. We would all like to <br />reach a definite conclusion. But where our total body of data is as limited in scope as pressures <br />of time and of money inevitably have made the weather modification experiments the world has <br />so far seen, dilution of our data by the asking of multiple questions -- an inevitable part of an <br />effective exploration (whether or not by separation) -- is guaranteed to keep us from being able <br />to draw firni conclusions. <br /> <br />Varying degrees of encouragement, hard to assess and assessable only through judgment <br />rather than formalism, are the best outcomes we can expect from after-the-fact exploration. <br />We should recognize this fact and, equally, should recognize the real value of encouragement <br />whenever it is strong. For such encouragement is the only grounds we have had on whioh to <br />base the establishment ,of confirmatory phases. With the unlikely exception of a major break- <br />through in both detailed and overall understanding of the physical processes uniting c:louds and <br />rainfall, encouragement from separation, usually after the fact, is still the only likely basis for <br />the selection of well-chosen confirmatory phases -- for any major steps we may manalge to take <br />toward understanding and utilizing weather modification. <br /> <br />* ongoing modifi(;ations * <br /> <br />In an exploratory phase it is unreasonable to expect complete consistency between the <br />plan of one season and the plan of another. Experimenters will learn as they proc,eed -- and <br />some of what thi~y "learn" will be in fact "true". What we have not seen, and what we think <br />ought to be both generally available and frequently consulted, is a quantitative analysis of the <br />consequences of adoptiQg any of the following strategies, when a somewhat better plan becomes <br />available: <br /> <br />1) forgetting the past, and starting again (zero weight on past data). <br />