“…Figure 5 shows the event size distributions in various windows of 1000 test vectors and their approximation by the model curves l= k . The optimum values of k in each case is found to vary slightly so we replace k by kt = a + b:t (4) Normalizing the function, we get the expression for event-size probability P r o b ft = = t = 1 = 1= kt :1= P mt 1 1= kt ; 1 mt (5) As observed earlier in this section, the variance of the fallout, mt, decreases quickly as the test progresses. In this work we have modeled mt in the same way as the event probability so mt = m 0 :1 , 1 , e ,gt h where kt are mt are given by Equations 4 and 6 respectively.…”