Connected issues: Attribute Accuracy and Completeness

Two Parts:



Completeness

The question has always been where does it fit...

Examining the actual text:

4.5 Completeness

The quality report must include information about selection criteria, definitions used and other relevant mapping rules. For example, geometric thresholds such as minimum area or minimum width must be reported.

This part links to lineage, in that it records the decisions applied. However, things like a minimum area or minimum width can be subjected to a test using internal evidence (checking the database: ARE things the sizes stated?).

In encoding cartographic features, standard geocodes (such as the feature codes described in Section 3.2 or in the FIPS codes for states, counties, municipalities and places) shall be employed as far as possible. Deviations from standard definitions and interpretations must be described.

SDTS promotes the use of certain other standards for coding. This is related to lineage in some degree, but it also relates to attribute accuracy. The SDTS feature set, as finally adopted is a subject of controversy, and will remain so. Completeness for political units (FIPS-coded counties) is quite different from 'rivers' or 'lighthouses'...

The issue of measurement framework is hidden in this text. Isolated objects have a different problem with completeness compared to exhaustive coverages. 'Spatial exhaustiveness' in the next paragraph applies to exhustive coverages, not to isolated objects. Spatial exhaustiveness involves gaps and overlaps ­ important issues for data quality...

The report on completeness shall describe the relationship between the objects represented and the abstract universe of all such objects. In particular, the report shall describe the exhaustiveness of a set of features. Exhaustiveness concerns spatial and taxonomic (attribute) properties, both of which can be tested. A test for spatial completeness can be obtained from topological tests for logical consistency that respond to the questions in 4.4. Tests for taxonomic completeness operate by comparison of a master list of geocodes to the codes actually appearing in the file. The procedures used for testing, and the results, shall be described in the quality report.

'Abstract universe', what is that? This text is written to handle a test where there is some external list of all the object that should appear. This phrase has been used as a (partial) translation of <terrain nominale>.


One important example involves parcel databases. The geometric descriptions of parcels should link 1 to 1 into the parcel attribute tables. This is often more complex than it seems.
Test of parcel completeness: link parcel id in geometric description (often Public Works or Engineering) to 'same' key in the attribute database (assessment or planning)
Report geometric objects not present in attribute table and reverse. (# or %)
Resolve these differences (some due to time; some due to definitions)