The data collected on the occasion of a cross-sectional epidemiological study on atherosclerosis (129 parameters for each of the 843 subjects) have been processed by means of an IBM 360/20, a computer not intended for scientific problems, and lacking magnetic storage media. In this purpose, it has been necessary to program down to the bit level and to set up a system allowing automatic feeding of the data and serial linking of the calculations, in order to work out all the dichotomies needed in the establishment of the diagnosis.
A pilot cross-sectional study on atherosclerosis has been carried out at the "Centre Hospitalier" of Castres, France. Data have been collected for a population sample of 843 subjects. Among these 843 subjects, there were 540 inpatients (all adults admitted to the hospital during a period of four months) and 303 people, more or less healthy, having a routine checkup for industrial medicine.
Each of the subjects has been assigned one of the three characteristics: Normal (N), "Pathological" (P), or Suspect (S). We define as suspect an individual who appears both not normal and not definitely pathological. Thus, the characteristics N, S, and Pcreate a partition in the population sample, and we look for the significant criteria of such a partition. We also may be brought to draw conclusions on the complementary ensembles, formed by: the not normals, N* = P>, the not suspects, S. = NJ, the not pathological, P. = E.
Data have been processed on a IBM card computer 360/20. Since this type of computer is not suited for scientific calculations, it has been necessary to overcome a certain number of diflicuhies.