The use of an electronic computer in the estimation of sampling errors in a nutritional survey.

If observations covering the whole of a body of material are available, the errors which are to be expected when sampling further material of the same type can be estimated and the relative accuracy of different sampling methods determined. Unfortunately, however, the numerical work required for investigations of this kind is very considerable, and in consequence when only desk machines are available statisticians are often reluctant to undertake this type of investigation with sufficient thoroughness to build up an adequate body of knowledge on sampling errors. The advent of electronic computers has completely changed the situation, for it is now possible to carry out the required numerical calculations without difficulty. Since work of this kind is essentially repetitive it is eminently suitable for electronic computers, for once a set of instructions (technically known as a programme) appropriate to a particular problem has been written for a machine, further data of the same type can be analysed with little effort. The present paper deals with the relatively simple but important problem of sampling numerical sequences. The investigation was made in order to determine the number and length of the periods over which an individual’s food intake should be observed in order to give sufficiently accurate assessment of the intake of the various dietary components. The nutritional aspects of the investigation and the results are reported by Chappell(1955). Since sampling problems of this type frequently arise in nutritional work and in other fields, it is considered that a brief description of the procedure adopted will be of value. Now that electronic methods of computation are available this type of calculation should become a standard routine when thorough knowledge of sampling errors is required for planning further work.