Setting Up the Frame - Sampling Spatial Units for Agricultural Surveys - page 92

In practice, statisticians have to deal with less than perfect situations, as several
limitations can affect the quality of the frames. Ignoring or failing to properly
identifying these limitations can lead to biased estimates, inflated variances, or even
both (Ferraz 2013 ).
Sometimes, mainly for budget reasons (it is very expensive building up a frame
for the whole population), a master sampling frame is a better solution (see Sect.
6.7 ). It is a classical solution to create the sampling frame from a census and/or
other administrative data (Carfagna and Carfagna 2010 ), for using in selecting
samples for different surveys or different rounds of a periodic survey. The aims
of the development of a master sampling frame are: avoiding duplication of efforts,
reducing statistics discrepancies, connecting various aspects of the phenomenon
allowing the analysis of the sampling units from the different viewpoints, and
having a better understanding of the phenomenon.
A master sample is a sample from which subsamples can be selected to serve the
needs of more than one survey or survey round (United Nations 1986 ), and it can
take several forms. A master sample with simple and rather common design is one
consisting of primary sampling units. The sample is then used for two-stage sample
selection (see Sect. 6.6 ), in which the second-stage sampling units are usually
spatial units. The main objective should be to provide samples that have reasonably
compatible design requirements with respect to domains of analysis and the distri-
butions of their target populations within those areas.
The main purpose of a master sample is to provide samples in a continuous
survey program (and any ad hoc survey that fits into the master sample design). The
sample will thus primarily be designed to serve a basic set of household, businesses
(or other kinds of units) surveys. In some situations, it may be possible to use the
master sample for surveys concerned with the study of characteristics of economic
units, such as household enterprises, own-account businesses, and small-scale
agricultural holdings. The main advantage of a master sample is the cost saving,
since the expenditures of selecting the master sample units and of preparing maps
and subsampling frames of target units will be shared by all the surveys using the
master sample. Furthermore, the setting up of a master sample is much cheaper than
making a frame of the entire population.
The layout of this chapter is the following. Section 5.2 is devoted to an overview
about the definition of the statistical units in sampling, describing the advantages
and the weaknesses of the sampling approach. Section 5.3 contains a review of the
strengths and weaknesses of statistical frames. Section 5.4 deals with the frame
construction. Finally, the last section concludes the chapter.
5.2 Choice of the Statistical Unit
The statistical unit for a data collection is the basic unit for which data are collected
(FAO 2005 ). Statistical units can be broadly divided in two main categories: legal
bodies and spatial units.