Completeness

1
Completeness

• February 27, 2006
• Geog 458 Map Sources and Errors

2
Outlines

• Completeness
• Testing completeness
• Documenting completeness in the metadata
• Data quality

3
Completeness

• The data set is called complete if whats
  defined/needed is encoded in the DB
• Spatial completeness degree to which all
  features are captured corresponding to data
  capture specifications
• Attribute completeness degree to which the
  relevant attributes of a feature are available
  corresponding to a given capture specifications
• Data quality component that describes whether the
  entity objects represent all entity instances of
  the corresponding abstract universe
• Relationship between the objects represented in
  the data set and the abstract universe of all
  such objects

4
Abstract universe

• Can be thought of a reference frame
• Data set digital representation of a subset of
  (perceived) reality
• Abstract universe terrain nominale abstract
  view of the universe universe of discourse
  miniworld subset of perceived reality (it
  involves selection and abstraction process)
• Data set is intended to represent the abstract
  universe
• Since completeness means the relationship between
  data set and abstract universe, a useful
  characterization of completeness relies on a
  comprehensive definition of the abstract universe

5
Data completeness vs. Model completeness

• It is possible to classify completeness into two
  categories depending on how the abstract universe
  is defined or specified
• Data completeness the abstract universe is
  defined on generic uses of data
  application-independent
• Model completeness the abstract universe is
  defined on specific uses of data
  application-dependent
• So which would be more flexible? Which would have
  multiple versions of completeness on the same
  data?

6
Spatial completeness

• Lets say the abstract universe lake is defined
  as the water body with the area more than 1
  square mile
• Check the number of entities in the abstract
  universe set this number to A
• Check the number of entities encoded in the DB
  (lake data set) set this number to B
• Completeness would be B/A
• The definition of lake varies depending on
  applications, thus so does A vary

7
Attribute completeness

• Subordinated to spatial completeness
• Define what the relevant attributes will be
• Lake will have area, depth, type (freshwater),
  and so on
• Check if attribute values are missing for entity
  in hand
• Geometric description might be incomplete (area)
• Report on the number of missing values out of the
  total number of features for each attribute

8
Relation to other data quality components

• Completeness may affect the logical consistency
  of a data set
• Missing arc, node ? connectivity, closed polygon
• Missing attribute (left and right-node) ?
  connectivity
• Missing attribute in PK ? key constraint
• Missing attribute in FK ? referential constraint
• So where do I document this in completeness or
  logical consistency?
• If incompleteness causes logical inconsistency,
  describe it in logical consistency section
• Else it will be included in completeness section

9
Data quality vs. fitness of use

• Data quality
• The totality of features and characteristics of a
  data set that bear on its ability to satisfy a
  stated set of requirements application-independen
  t
• Fitness of use
• The totality of features and characteristics of a
  data set that bear on its ability to satisfy a
  set of requirements given by the application
  application-dependent

10
Data quality vs. fitness of use

• Data quality information is usually provided by
  the producer of a data set
• Fitness of use is assessed when evaluating the
  use of a data set by users ? this principle is
  referred to truth in labelling (users are
  responsible for quality control indeed)
• See different approaches to quality control in
  the lecture note on spatial data quality

11
Data quality report

• What you are reporting in data quality section of
  the metadata will be data-independent, so that it
  can be reused for any potential uses of the data
• Reporting data quality can be thought of the
  process for evaluating the ability of the data
  set to meet up to the requirements
• In that how well the value is close to ground
  truth (attribute/positional accuracy), whether it
  exhibits lack of contradictions (logical
  consistency), and whether whats relevant is
  encoded in the DB (completeness)