№ 4, 2014
The paper presents a formalized approach to the detection of systematic errors in mass surveys by constructing a single data matrix. Errors are defined as the difference between values of individual variables in the sample representing all respondents and a subsample, from which units of observation coded under different contextual conditions or possessing another set of related parameters are removed. The time the interview was conducted and its duration are considered as factors that hypothetically affect biases. The results obtained by the authors do not allow to speak with confidence about the reliability of the developed procedure, but presents great opportunities for operationalization of the general theory of errors.
Main Page ~ Authors ~ Saponov Dmitry