Dummy Modeling Properties

CSVW-EO defines properties useful for generating realistic dummy datasets.

These properties improve dummy data quality while remaining compatible with differential privacy workflows.

Warning

If they disclose private information, they should not be added.

Nullable Proportion

Property	Meaning
nullableProportion	Approximate fraction of null values
maxNumPartitions	Maximum number of partition (keys)

Example:

{
  "nullableProportion": 0.15,
  "maxNumPartitions": 5,
  "publicKeys": ["January", "February"],
  "exhaustiveKeys": false
}

These column will be generated with arounf 15% of null values and take values in ["January", "February", "a", "b", "c"]. The public keys and then random keys to have the same number of partitions as maxNumPartitions.

Note: nullableProportion may be approximate.

Dependencies

Dependencies describe relationships between columns.

Property	Meaning
dependsOn	Source column
dependencyType	Type of dependency
valueMap	Mapping definition

Dependency Types

bigger

Indicates that one column is always greater than another. Only useful for columns whose bounds overlap.

Example:

• date_treatment_1 > date_treatment_2
• date_of_death > date_of_birth

mapping

Defines deterministic or constrained mappings.

Example:

{
  "dependencyType": "mapping",
  "valueMap": {
    "child": false,
    "adult": true
  }
}

Another example:

{
  "dependencyType": "mapping",
  "valueMap": {
    "medical": ["doctor", "nurse"],
    "engineer": ["civil", "mechanical"]
  }
}

fixedPerEntity

Indicates values remain constant for the same privacy unit.

Example:

• birth date per patient
• country of birth per individual
• height per (adult) person

Public Keys

CSVW-EO can define known public domains.

Property	Meaning
publicKeys	List of public values
exhaustiveKeys	Whether keys are exhaustive

Example:

{
  "publicKeys": ["January", "February"],
  "exhaustiveKeys": false
}

{
  "keyValues": [
    "January",
    "February",
    "March",
    "April",
    "May",
    "June",
    "July",
    "August",
    "September",
    "October",
    "November",
    "December"
  ],
  "exhaustiveKeys": true
}

If exhaustiveKeys=true, all possible keys are publicly listed.

Partitions

Partitions describe public regions of the domain.

Property	Meaning
partition	List of csvw-eo:Partition objects
exhaustivePartitions	Whether partitions fully cover the domain

Examples:

• categories
• numerical intervals
• grouped partitions

Partitions may be:

• exhaustive
• overlapping
• disjoint

Both csvw:Column and csvw:ColumnGroup may define partitions.

Example: Column Partitions

Example partitioning on the species column:

{
  "name": "species",
  "datatype": "string",
  "partitions": [
    {
      "@type": "Partition",
      "predicate": {
        "partitionValue": "Adelie"
      },
      "maxLength": 152
    },
    {
      "@type": "Partition",
      "predicate": {
        "partitionValue": "Gentoo"
      },
      "maxLength": 124
    }
  ],
  "keyValues": [
    "Adelie",
    "Gentoo"
  ],
  "exhaustiveKeys": true,
  "maxNumPartitions": 2
}

Example: Column Group Partitions

Example partitioning on the (species, island) column:

{
  "@type": "ColumnGroup",
  "columnsInGroup": [
    "species",
    "island"
  ],
  "partitions": [
    {
      "@type": "Partition",
      "predicate": {
        "species": {
          "partitionValue": "Adelie"
        },
        "island": {
          "partitionValue": "Dream"
        }
      },
      "maxLength": 100
    }
  ],
  "maxNumPartitions": 5
}

Column groups describe partitions resulting from grouping on multiple columns simultaneously.