|
<br />..
<br />
<br />90a
<br />
<br />Dr. Braham emphasizes the problems that arise from
<br />lack of a strong, accurate, easily applied model, and from
<br />problems of execution of the experiments. If the most
<br />earefully controlled experiments face difficulties in execu-
<br />tion and interpretation, how can we expect ro get much
<br />information from nonrandomized commercial studies as
<br />suggested by Dr. Braham at the close of Section 51 Dr.
<br />Braham tells us that the current physical theory is not
<br />adequate, but such theory is usually the basis for
<br />strengthening nonrandomized analyses. Possibly these
<br />studies contain information of a different kind that would
<br />be useful for policy. For example, they may suggest con-
<br />straints that should be put on two or more commercial
<br />enterprises operating in the same neighborhood or region,
<br />where the enterprises might nullify one another's efforts
<br />or create effects different from those desired. Or perhaps
<br />these studies co:uld provide suggestions for organization
<br />or reporting. At the conference mentioned previously, the
<br />members discussed the preservation and exchange of
<br />data. They suggested that the feasibility of creating a
<br />central reposirory for data from weather modification
<br />experiments be explored. The Panel went further and
<br />recommended that
<br />
<br />NOAA [National Oceanic and Atmospheric Administration]
<br />create a Repository for Data on Weather-Modification Activi-
<br />
<br />JERZY NEYMAN*
<br />
<br />Journal of the American Statistical Association, March 1979
<br />
<br />ties" which, at a suitable price, will make available for re-
<br />analysis complete data on these activities (Committee on
<br />Atmospheric Sciences 1973, p. 198).
<br />
<br />With respect ro disagreements about reanalyses (men-
<br />
<br />tioned in the third point of Braham's summary), the
<br />
<br />Panel said (Committee on Atmosphelic Sciences 1973,
<br />
<br />p. 198) :
<br />
<br />We anticipate that there may be substantial disagreements
<br />about the outcomes of such reanalyses. In spite of this, argu-
<br />mellts between scholars who have made deep analyses are likely
<br />to IDe more productive than among those making superficial
<br />analyses.
<br />
<br />REFERENCES
<br />
<br />Changnon, A., Jr., Farhar, Barbara C., and Swanson, Earl R. (1978),
<br />"Hail Suppression and Society," Science, 200, 387-394.
<br />Committee on Atmospheric Sciences (1973), Weather and Climate
<br />ModiJ'icaticm: Problerm and Progress, Washington, D.C.: National
<br />Research Council, National Academy of Sciences.
<br />Fairley, William, and Mosteller, Frederick (1977), Statiatica and
<br />Publu: Policy, Reading, Mass.: Addison-Wesley Publishing Co.
<br />Howard, Ronald, Matheson, James E., and North, D. Warner (1972),
<br />"The Decision to Seed Hunicanes," Science, 176, 1191-1202;
<br />followed by Power, Bernard A., and Kates, Robert W. (1973),
<br />"Letters to the Editor on Seeding Hurricanes," with reply by
<br />Howard, Matheson, and North, Science, 179, 744-747; followed
<br />by Sundqvist, Hilding (1973), "Letter to the Editor on Hunicane
<br />Seeding Analysis," and reply by Howard, Matheson, and North,
<br />Scienl:e, 181, 1072-1073.
<br />
<br />Comment
<br />
<br />.
<br />
<br />In his nicely formulated article, Professor Braham calls
<br />for closer contacts and cooperation between meteorol-
<br />ogists engaged in weather modification, on the one hand, .
<br />and statisticians on the other. On this point we are in full
<br />agreement. I also agree with Braham that the study of
<br />effects of cloud seeding is useful for mathematical sta-
<br />tistics. As with any complex natural phenomenon, efforts
<br />ro evaluate the effects of cloud seeding generate in-
<br />teresting problems in the theory of statistical tests. An-
<br />other important point of agreement is Braham's assertion
<br />that "theories of cloud response ro seeding ... are
<br />simplistic and inadequate."
<br />However, we do have important disagreements. The
<br />following remarks are formulated with reference ro Pro-
<br />fessor Braham's somewhat emphatic regrets about
<br />
<br />· Jerzy Neyman is Director of the Statistical Laboratory, Depart-
<br />ment of Statistics, University of California, Berkeley, CA 94720.
<br />This discussion was prepared with the partial support of the Office of
<br />Naval Research, Contract No. ONR-NOOOI4-75-C-OI59, and of the
<br />U.S. Army Research Office, Grant No. DA AG 29 76 G 0167. All the
<br />opinions expressed in this discussion are those of the author.
<br />
<br />
<br />"Neyman and his associates in Berkeley." These remarks
<br />will include Braham's Tables 1 and 2.
<br />The principal point of disagreement between Professor
<br />Braham and myself is his philosophical attitude expressed
<br />in the statement: .
<br />
<br />As physical scientists, we naturally think in terms of physical
<br />models. We are inclined to accept evidence which seems com-
<br />patible with established meteorological physics, even though it
<br />ma,y have modest statistical support, over evidence that runs
<br />counter to accepted physics, even though the latter may have
<br />stl"onger "statistical" support.
<br />
<br />As a research worker interested in understanding the
<br />atmospheric phenomena connected with cloud seeding, I
<br />am skeptical of any preconceived idea (e.g., "established
<br />meteorological physics"). This skepticism is supported by
<br />Braham's own assertion just quoted, "theories of cloud
<br />response ... are simplistic and inadequate." My phi-
<br />losophy is: (i) to find out what the fact,.c; are (which is the
<br />"explorarory" phase of the study), (ii) to verify the
<br />generi~lity of these facts (confirmarory phase), and, in the
<br />
<br />
|