Epidemiological Field Study In an early field study in Hidalgo County, Tex. (4), simple indexes were used to deter-mine whether fly control in a poorly sanitated area would lower the incidence of diarrhea and enteritis. A group of small communities was divided into two sets. For a period of several months, April 1946 through August 1947, the lfies were controlled in one group but not in the other. Control was accomplished by application of residual insecticide every 6 weeks, with spot re-treatment when necessary. At that time, development of resistance of flies to DDT had not become a problem. In September 1947,mlfy control was discontinued in the first group of communities and begun in the second group. In each group a fixed set of blocks was examined for flies each week. The three highest grill counts in each block were recorded, and the highest reading was used as a statistic for comparison with data on diarrheal disease incidence collected from residents of the area. From these area residents, data on incidence of diarrhea was obtained through household visits by interviewers. The infection index chosen was the attack rate among children under 10 years old, and from this group monthly samples of stools were cultured for Shigella and Salmonella. A summary of results on fly prevalence and incidence of diarrhea in the treated and untreated communities is presented in figures 2 and 3. The many sources of variability in individual measurements apparently did not impede recognition of the main effects of the treatment. Interpretation of Results In some entomological surveys, there may be no comparable control population from which a random sample may be selected. This point may be illustrated by entomological appraisal of the effectiveness of treatment of households with residual insecticide during the final stages of malaria eradication in the United States. The counties selected for treatment were those which in the past had had the highest malaria rates. In these counties, densely inhabited areas with the heaviest A. quadrimaculatus breeding potential were given preference for treatment. Hence, there was no comparable control area. Selection of individual comparable houses in the treated areas and in the adjoining untreated areas was left to the discretion of the entomologist.The index chosen was a simple onepresence of live A. quadrimaculatus in a home during the afternoon. Movement of these anophelines during the daytime is negligible so that it could be assumed that any found in homes during the afternoon had entered not later than the preceding night. Because the residual insecticide was 'slow in action, no inspections were made during the morning hours.In the
course of 4 years of work, approximately 47,000 treated and 5,000 untreated houses were inspected (5) ; live A. quadrimaculatus were found in 1.7 percent of the treated and in 16.0 percent of the untreated houses.The difference between the two percentages is manifestly significant by any statistical test. However, interpretation of the significance depends upon one's judgment of the ability of the entomologist to select comparable samples of treated and untreated houses. Although this is an extreme example, it is illustrative of the qualitative judgments which enter into most entomological sampling problems.Results of entomological surveys are influenced by many variables which cannot be controlled and by many subjective factors not amenable to routine measurement. Statistics used to describe the findings should be simple and readily grasped by the epidemiologist, the entomologist, and the engineer. By using simple statistics, qualitative interpretations may be made jointly by the statistician and his colleagues in the natural sciences. Interpretation of elaborate statistical methodology depends in many instances on the statistician's intuitive appraisal of the extent to which theoretical assumptions may be violated. On these questions the natural scientist usually has less basis for making judgments, and he may regard the methodology as an "ivory tower," and of questionable value. Therefore, the statistical presentation should be in terms of statistics whose import is fully understood by all members of the group.Many phases of an entomological survey may give rise to special problems for which planned experiments of complex statistical design are necessary. Among these are tests of collecting devices, individual differences among collectors, and design of subsampling methods for rapid estimation of the composition of trap collections.However, even in the planned entomological experiment, the intangibles of skill, effort, and thoroughness play a leading role, and the statistician must learn to reckon with these factors as familiarly as with the tools of his statistical methodology