WMOD00464 - Laserfiche WebLink

Home Browse Search

A-4 we recognize it, in these particular circumstances, as a special public need ~ '" data management * Good data not only requires careful supervision through its first recording; it requires careful and thoughtful handling thereafter. If, as is becoming more economical (and often necessary if we are to do a good job), it is to be entered into a computer system, the quality (in maintaining appropriate tags and avoiding undesired losses of data, for instance) of the data management system involved deserves careful attention. Indeed, the whole plan of data management requires careful thought before data collection is begun, and a careful re- examination at suitable intervals. '" cost and value * To improve data quality will increase costs. Will the return be worth it? Poor data can, and has in the past, made the results of experiments nearly useless. Under such circumstances nearly 100% of the experiments' costs were wasted. In our judgment it is likely to be worthwhile to increase the cost of individual experiments by up to 20% or 25% in order to assure quality data, even if, as is almost certain, this causes a corresponding reduction in the rate at which experiments can be conducted. Good data is worthwhile, poor data is not. 3. Structure of Experimentation * purposes of exploratory phases * The distinction between exploratory and confirmatory phases is a recent one in the field of weather modification. There does not yet seem to have been time for an adequate understand- ing to develop of the breadth of purposes appropriate to an exploratory phase; so far, an exploratory phase seems only to have been thought of as an earlier version of the confirmatory phase, e.g., one where it is appropriate to look for more restricted situations where seeding will be more rewarding. This is an important end, fully worth careful distinction of exploratory from confirmatory; it is not the only purpose that should be recognized. As suggested above, one important aspect of exploratory phases may be learning how to approximate expert judgments by objective criteria. Another important purpose can -- and should -- be learning how to better measure the effects to be studied. This might involve the development of a taxonomy of storms, as seen in the area, or the identification of better covariates (better concomitant observations) or of better ways to use either new or old covariates. If our aim is reliable confirmation, as soon as' feasible, it is just as important, in such a difficult area as weather modification, to learn how to measure better as it is to learn how to pick favorable situations in which to attempt confirmation. * exploration by separation * The main step in the exploratory, after-the-fact analysis of weather modification experi- ments based on seeding is referred to by many meteorologists as "stratification"; statisticians might prefer to speak of "categorization"; philosophers or historians of science would presum- ably have their own preference. We will call it "separation". Whatever the term, the process 'SUCh policies ought, if adopted, help to make both blatant and useless such reporting practices as saying only "it is unlikely that seeding reduced hail by as much as 65%" when one could equally well add "or increased it by as much as 300%".