Some effects of selection on the JVE data

The threshold test ban treaty (TTBT) between the United States and the Soviet Union stipulates that all nuclear tests shall be below ground and shall not exceed 150 kilotons in yield. In order to assist both countries in monitoring the TTBT an historic experiment has recently taken place. In particular, a team of Soviet scientists and military personnel have conducted a nuclear test at the US test site in Nevada and a similar US team has conducted such a test at the Soviet semipalatinsk test site. One of the major purposes of these tests was to aid each country in calibrating their seismic monitoring equipment. To aid further with the calibration problem, both sides agreed to furnish the other with the yields, as measured at their respective sites, of five of their previous nuclear tests. This experiment is referred to as the joint verification experiment (JVE). Unfortunately, there is a major problem with the use of these data since they are unverified. That is, the data could be falsified or at least be selective in nature, i.e., the data may have been selected to intentionally mislead rather than being randomly chosen. This paper considers the effects of purposeful selection, rathermore » than random selection, on estimates of the mean and variance. Surprisingly, it is shown that if the selection is done in a manner that would most benefit the Soviets, then unbiased estimates for the mean and variance of the parent population of interest can still be obtained.« less