The problem involved in determining the number of cases necessary in the present study to provide an adequate sample of any larger population, has already been discussed in Chapter 3. The generalizations made there have been based upon a series of pragmatic tests of the values of calculations derived from samples of various sizes. The specific data derived from those tests are shown in the tables presented in this Appendix.
As previously explained (Chapter 3), calculations have been made systematically for samples of 50, 100, 200, 300, 400, and (where the material is available) 600, 1000, and still larger numbers of cases. In all, 698 different populations have been used in these tests. Each population has been homogeneous for sex, race, marital status, age, educational level, and either the rural-urban background or the religious background of the included individuals. The samples of various size have all been selected by a strict randomization performed on an IBM sorter; and in each instance the sample has been drawn directly from the total number of histories available in each population. In no case has a larger sample been built up by adding cases to an originally smaller sample of 50, 100, or other size.
In the tables given here, the statistics which are shown in boldface represent calculations which lie within 5 per cent, plus or minus, of the calculations obtained from the largest sample available in that group; except that comparisons of incidence data have been made with an allowance of 2 per cent, plus or minus. The boldface figures therefore represent results that are adequate, as judged by the results obtained from the use of the largest sample, and within the allowed range of error. Of course, an error of any other size might have been arbitrarily selected as a basis for judging adequacy in these tables. See Chapter 3 (Table 2) for a summary tabulation of the number of adequate samples which are shown here in these tables.
The eight tables which follow present the following statistics on each of the sample populations:
Table 155: Mean frequencies per week (and the standard deviation) on the total population in each sample
Table 156: Mean frequencies per week (and the standard deviation) for the active population in each sample
Table 157: Median frequencies per week for the total population in each sample
Table 158: Median frequencies per week for the active population in each sample
Table 159: Incidence of active cases in each sample
Table 160: Locus of the mode for the frequency distribution in each sample
Table 161: Height of the mode for the frequency distribution in each sample
Table 162: Range of variation (represented as the maximum frequency minus one case) in each sample
Wherever all of the medians in an entire section of any table have fallen into a zero frequency class, they have not been shown in these tables.
In column one, in each of the following tables, these abbreviations have been used:
Table 155. Size of sample necessary to secure stable means on total population
Boldface figures designate calculations that lie within 5 per cent, plus or minus, of the calculations obtained from the largest sample. Means are frequencies per week.
Table 156. Size of sample necessary to secure stable means on active population
Boldface figures designate calculations that lie within 5 per cent, plus or minus, of the calculations obtained from the largest sample. Means are frequencies per week.
Table 157. Size of sample necessary to secure stable medians on total population
Boldface figures designate calculations that lie within 5 per cent, plus or minus, of the calculations obtained from the largest sample. Means are frequencies per week.
Table 158. Size of sample necessary to secure stable medians on active population
Boldface figures designate calculations that lie within 5 per cent, plus or minus, of the calculations obtained from the largest sample. Means are frequencies per week.
Table 159. Size of sample necessary to secure stable incidence data
Boldface figures designate calculations that lie within 2 per cent, plus or minus, of the calculations obtained from the largest sample.
Table 160. Size of sample necessary to fix the locus of the mode
Boldface figures designate calculations that lie within 5 per cent, plus or minus, of the calculations obtained from the largest sample.
