Page 98 - Special Topic Session (STS) - Volume 1

STS407 Étienne Saint-Pierre
- The development of agency-wide measuring quality of estimates that
come from non-direct data sources (administrative data, satellite
images, models, etc.) with methodologically acceptable approaches.
- The introduction of innovative approaches (ex. random tabular
adjustment) to manage disclosure avoidance other than data
suppression to enable a greater amount of data to be published, but
without the need for complicated data perturbation of microdata
records.
- The introduction of improved methods of data collection and case
prioritization: The iterative derivation of a set of quality indicators to
support actively the management of collection and follow-up
activities is a new procedure replacing the more subjective static
methods of the past to identify what records should be given priority
for follow-up, when collection should cease in certain domains or
even overall.
- A more standardized approach to pre-processing data prior to record
linkage to ensure that the data sources that need to be linked are
handled in a consistent and high-quality manner, thus improving the
odds for successful linkages. New research in ways that record linkage
can be undertaken without requiring manual verification, but still
control the level of false results is being undertaken.
- The Corporate initiatives which can be used to coordinate work and
knowledge transfer in machine learning, artificial intelligence among
different parts of the agency. This also ensure a set of approved
common methods, enhancing the coherence of the Census of
Agriculture with other statistical programs.
- In addition, the use of common tools and databases facilitates the
transfer and sharing of resources between programs. There are
currently around 750 users of the IBSP tools and systems at Statistics
Canada. Given the cyclical nature of the Census of Agriculture and the
high demand of resources for a short period of time, this is an
important element to ensure well trained staff is available when
needed.

5. Discussion and Conclusion
The Census of Agriculture Program must adopt innovative, state-of-the-art
methodological and operational approaches supporting the integration of
multiple alternative data sources in its business model to take advantage of
the new data landscape. It is essential to respond adequately and in a timely
manner to data user needs. The use of the traditional approach of
administrating a questionnaire to every member of the farm population to
address the growing information needs is obsolete and has to be reviewed.

87 | I S I W S C 2 0 1 9