| Sumario: | The estimation of expenditures and other social variables that require relatively long periods of time for their determination and application, are particularly subject to four types of errors. Errors due to incompleteness of the subjects recall over preceding periods of time; errors due to incomplete recording by subjects in keeping records over long periods of time; errors due to the use of estimates for one period of time to apply to later periods of time; and errors due to the use of parameters estimated for the total population for the prediction for individuals within the population. The purpose of the present article is to demonstrate an application of statistical methods of sampling and estimation to control such errors. The first type of error was controlled by the use of a "recording" method instead of a "recall" method for enumerating expenditures of individuals. The second type of error was controlled by using sub-periods of one week, instead of the entire period of one quarter, as the period for which records were to be kept. The third type of error was controlled by using all the weekly sub-periods as "strata" within which random samples of individuals were selected to keep records covering the week for which they were chosen.
|
|---|