Surveys and decisions in the context of multidisciplinary programmes:: estimators and indicators

Abstract

Scientific programmes whose objectives are to provide pertinent knowledge and information for sustainable use of a natural resource always include data collection operations. Functions of the data collected are obtained from surveys. These functions (estimators) are formulae constructed according to the survey’s model defined by the design of the survey and the sample selection procedures. In complex designs, a ‘superpopulation model’ is always present. This model, accounting for available knowledge on the observed system, is defined by assumptions on the distribution of the collected data. Users consider not only the quality of the estimates (outputs of the survey) but also, and possibly primarily, the quality of these estimates as input (information) to their own decision frameworks, which also constitute a superpopulation model. Hence, a survey combines at least two models; a survey model and a user’s model. These different models are discussed in this paper with the help of three examples that differ in the types of system under study, in the objectives of the surveys, and in the nature of the collected information. Specific functions of the collected survey data may be considered as estimators when considered as output of the survey model and as indicators when considered as input to the user’s model. A user may build indicators from many sources of information and, as several different users may use the estimates, a survey is an element of an information system. Analysis of data leads to improved knowledge of the system under study and to the better identification of the various associated models. Therefore, surveys are elements of dynamic systems in which theory and practice are complementary. Although classical sampling theory often appears not to be sufficient in this global context, it remains necessary. The use of information from different sources to be used from different points of view implies an increased level of rigour with many models, all of which must be explicitly defined.