Laserfiche WebLink
<br />where <br /> <br />y * = the predicted unit area rainfall variable <br /> <br />Xl = the treatment (seeding) factor of interest, and <br /> <br />Xi (i = 2, ..., k) = the (k-1) predictor or concomitant variables. <br /> <br />The xi variable is different from the others because it is an index variable that signifies the <br />presence or absence of the randomized treatment and hence is termed a factor. This factor <br />will be examined for evidence of treatment (i.e., seeding effect). The other variables are <br />concomitant or predictor variables and are present to account for some of the natural <br />variability in the rainfall response variable (y). <br /> <br />The concomitant or predictor variables to be selected for inclusion in the linear model will <br />probably come from the categories of meteorological processes that likely influence rainfall <br />in Thailand. These categories include: <br /> <br />1. Mesoscale prewetness as estimated by radar-measured rainfall rates in the target or <br />adjacent areas prior to the time of first unit treatment, <br /> <br />2. Thermodynamic structure ofthe atmosphere as measured by the GPCM (Great Plains <br />Cumulus Model) and/or by standard stability indices, <br /> <br />3. Kinematic structure of the atmosphere as represented by wind speed and/or direction <br />and vertical wind shear. <br /> <br />4. Cloud-scale suitability of the day as indicated by convection having substantial SLVV'C <br />(i.e., > 1.0 gram per cubic meter) and strong updrafts (i.e., > 5 meters per second), and <br /> <br />5. Synoptic scale rainfall as estimated by radar reflectivity for large areas that are some <br />distances from the target and hence presumably unaffected by treatment. <br /> <br />The candidate predictor variables to be selected from the above five categories will be <br />identified during the course of the project in consultation with Thai project scientists. <br /> <br />2.11.4 Statistical significance tests <br /> <br />The probability that the results of the demonstration project are caused by chance will be <br />calculated by a refinement of the Monte Carlo rerandomization test (Gabriel and Feder, <br />1969). This calculation will be done in the following steps for the experimental units: <br /> <br />1. A randomization reallocation of the experimental random cases to Sand NS will be <br />made. If the randomization is constrained to disallow runs of the same treatment <br />decision within each block, the rerandomization draws will be similarly constrained. <br /> <br />2. The SR between all the Sand NS randomized cases will be calculated. <br /> <br />20 <br />