Laserfiche WebLink
<br />2.11.3 Accounting for natural biases <br /> <br />2.11.3.1 The identification of control cells <br /> <br />Any analysis of seeding effect must indude an accounting for at least some of the natural <br />variability. Our first possible approach will be that pioneered by Gagin et al. (1986) and <br />refined by Rosenfeld and Woodley (1989), which made use of "control" cells in the <br />environments of the AgI-treated and simulated AgI-treated cells. This approach was not <br />feasible, however, for the experimental units obtained in Texas in 1989 and 1990, because <br />too few control cells could be identified on some days of experimentation. The use of this <br />approach in Thailand, therefore, will depend on whether a sufficient number of control cells <br />can be identified on each day of experimentation. This approach is described in detail below <br />because it may be used in Thailand. <br /> <br />If the properties of control cells, as defined momentarily, were perfectly correlated with the <br />cells within a NS experimental unit, then the seeding effect could be determined by the DR <br />(double ratio): <br /> <br />DR = SRICR <br /> <br />where SR (single ratio) is the simple apparent seeding effect, given by: <br /> <br />SR = ( )T Ag/( )Tsim <br /> <br />and CR (control ratio) is the bias causEld by the natural variability, given by: <br /> <br />CR = ( )C AgI( )Csim <br /> <br />where the values in parentheses can be the mean of any ofthe lifetime properties of the cells; <br />T stands for treated cells, C for control cells; AgI denotes the cells in which AgI seeding took <br />place, and "sim" denotes the cells in which seeding was simulated. <br /> <br />., <br /> <br />In defining the cells to be used as controls, Rosenfeld and Woodley (1989) had to consider <br />several factors. First, the prospective cells had to conform as much as possible to the <br />selection criteria of the actual experimental cells that received S or NS (simulated) treatment. <br />Second, the control cells had to be separated from the S and NS cells by a minimum distance <br />to ensure that the experimental cells did not contaminate those cells, which were to be used <br />as controls. Third, because of range biases in cell measurements that can result from the <br />characteristics of the measurement radar, each set of control cells had to be located as far <br />from the radar as the experimental cellls that resided within each experimental unit. <br /> <br />The initial criteria for the selection of the control cells in the AARRP will be the following: <br /> <br />1. The prospective cell must be tracked for at least two radar scans, <br /> <br />2. The cell is no more than 125 kilometers from the radar at the time of its birth, <br /> <br />3. The prospective cell is never within 35 kilometers of any treated cell, <br /> <br />4. The height of the prospective control cell must be at least 6.5 kilometers on the second <br />radar scan, and it must be taller on the second scan than on the first, but not more <br />than 10 kilometers, <br /> <br />17 <br /> <br />~ <br />