Survey Data Collection and Processing - Sampling Spatial Units for Agricultural Surveys - page 232

Accuracies of individual land cover/land use classes may be also assessed. Story
and Congalton ( 1986 ) distinguished between the producer s accuracy and the

'

user s accuracy of the land cover/land use class expresses

the conditional probability that a randomly selected unit classified as category i by
the reference data is classified as category i by the survey. It is referred to as
producer

s accuracy . The producer

' '

s accuracy because the producer of a land cover/land use survey is
interested in how well a reference category is depicted in the data. The user

'

s
accuracy for land cover/land use class i expresses the conditional probability that a
randomly selected unit classified as category i in the survey is classified as category
i by the reference data. The row and column totals can also be used to quantify the
probabilities of omission and commission errors.
Furthermore, most accuracy measures can be estimated by the HT estimator
because they are expressed as totals (Stehman 2001 ). Specific guidelines for
implementing consistent estimators for accuracy parameters were given by Strahler
et al. ( 2006 ). An approach for estimating the variance associated with estimated
accuracy measures was discussed by Stehman ( 1995 ), and a general formula for the
variance estimator was proposed by Strahler et al. ( 2006 ).
However, we can define some alternatives to these direct estimation measures of
quality that make use of indicators of the sampling process. These statistics are
cheaper, because they are based on the available information, and are examples of
an indirect estimation.
This indirect approach allows us to monitor each step of the data production
process, and to organize corrective action plans if the indicators suggest that there
are problems.

'

Conclusions
This chapter revealed how data collection, data editing, and quality assurance
operations could be organized and executed, with various levels of technical
details. We were particularly interested in the organizational aspects and the
management of data flow from one stage of a survey to another. Our focus is
on the ability of all personnel involved, particularly the enumerators or
interviewers. Their training has a critical role in the correct implementation
of a system that can produce timely and reliable basic data, which can then be
used as the foundation of high quality estimates. We have addressed public
relations issues, including survey publicity campaigns, respondent relations,
and general public relations. We have discussed some technical practicalities
of data editing such as procedure characteristics and features, sequences of
operations, error detection and identification, and data imputation techniques.
Finally, we examined quality and performance monitoring methods.
To ensure that the survey data collected are complete and accurate, decisive
attention should be paid to any warning from either a manager or employee
involved in the project. From this point of view, the organizational structure
should be sufficiently flexible to allow the survey to adapt to new conditions
encountered in the field and to new problems not foreseen in the design phase.