Laserfiche WebLink
<br />2. Stratification. - Because no powerful control variables have <br />been identified to predict the behavior of the individual test <br />clouds, a pretreatment stratification of the clouds into classes <br />(A-1, A-2, and B) will be used in HIPLEX-1. This stratification <br />is based on knowledge gained from the background studies in <br />Miles City since 1975. The physical criteria are listed in <br />section VII. <br /> <br />Not all the data sets of HIPLEX-1 will be of uniform quality. The <br />"best" data sets will be gathered from those events for which the <br />Miles City radar observes the +10 oC level and for which all <br />instrument packages operate properly and the aircrews and ground <br />crews execute the test procedures correctly and on time. Occasion- <br />ally, a variable will not be measured because of such events as <br />instrument failure during a mission. However, once the randomiza- <br />tion envelopes are opened, the case becomes part of the a priori <br />data set, regardless of what happens subsequently. <br /> <br />Stratification by quality of the data sets will be included in the <br />statistical evaluation and will result in subsets of events <br />degraded by the absence of one or more response variables. Fer <br />example, during a giVE!n season 40 cases might be selected, 30 of <br />which contain adequate measurements of all 13 primary response <br />variables to be examined in the statistical evalui!tion. Six cases <br />might be missing the decelerator slides from which CCR5 is deter- <br />mined, and another four cases might be missing the liquid water <br />content device from which AWC8 is determined. The following <br />subsets could then be used: <br /> <br />a. Thirty complete! events (primary data set for evaluation). <br /> <br />b. Thirty-four events that consider 12 response variables, <br />with AWC8 excluded (30 complete events with AWC8 deleted and <br />the 4 cases with AWC8 missing but CCR5 present). <br /> <br />c. Thirty-six events that consider 12 response variables, <br />with CCR5 excluded (30 complete cases with CCR5 deleted and <br />the 6 cases with CCR5 missing but AWC8 present). <br /> <br />d. Forty events that consider 11 response variables, with both <br />CCR5 and AWC8 excluded (30 complete cases with both CCR5 and <br />AWC8 deleted, 6 cases with CCR5 missing and AWC8 deleted, and <br />4 cases with CCR5 deleted and AWC8 missing). <br /> <br />The stratification is more severe for the radar-derived response <br />variables. Radar datcl are not recorded within 25 km of the SWR-75 <br />radar (range delay), clnd beyond about 85 km there is no assurance <br /> <br />29 <br />