The unit problem in official business statistics

(A) Presenter

Boris Lorenc, Statistics Sweden

(B) Category

Early-stage idea

(C) Partners needed

Researchers (academics) in the fields of mathematical statistics; experts in formalisation of properties and relations of units (some formal logic machinery)


There is a considerably large variety of units and their relations in business statistics. The unit error – that is, the error associated with incorrect specification or characterisation of business units in the sampling frame – has thus far received fairly little research attention. A number of arguments for a need to bring the unit identification problem under increased scrutiny of survey methodologists include:
(i) consistency with methodology of other efforts towards quality improvement;
(ii) establishment of a fact based decision making process regarding statistical units;
(iii) integration of unit uncertainty into general survey methodology theory and practice.
A number of specific areas suitable for research will be presented, embedded in two different approaches to the unit problem.