4 sashbsinntio ta, é CHAPTER 6 SPECIAL STUDIES 6.1 VARTARTLITY = OF SIMTETODOS GuttmD PAPERS — L. Dhicicldtih, ta echtboanbdacie The current state of knowledge on collection characteristics of gummed paper leaves much to be desired. An important problem ‘in connection with gummed paper samples concerns the representativeness of individual collections. Data for investigation of ‘ona phase of this problem was obtained by exposing pairs and triplets of gummed papers simultaneously about six feet apart during some of the test operations... The basic data, unfortumately, are not amenable to standard ‘statistical tools available tg test the significance of differences between simultaneous samples. Analysis of the problem indicates ‘that we must make additional assumptions about the average radio“activity per particle as well as the constancy of that average; or abandon the. attempt to obtain absolute results and be satisfied with comparisons. Only the second alternative appears justified ' at this time. A straight-forward approach could be based on two assumptions; a) fallout on papers exposed six feet apart is from a homogeneous ~ cloud of radioactive particles, i.e., no gradient of particles over a distance of six feet, and b) collection efficiency of simultaneous paper is identical. As a.consequence of the assumptions one would expect identical collections and the observed departures from equal collections could be examined for significance, A serious difficulty arises, however, because the tests of significance depends upon the number of particles collected on each paper, not on some quantity proportional to that number. For example, suppose the average radioactivity collected on two given pairs is such that 0% of the total activity is on one paper. Were this activity caused by 10 particles per pair, a division as different. from 5-5 as is l-6 has approximately a 0.5 probability of occurring by chance. On the other hand, if the activity were due to an average of 1000 particles per pair, an average division as different from 500-500 as is 00-600 has a probability of less than 0.01 of being equalled or exceeded purely by chance. ~ 2 -