Add categoric data to qdata:Observation

[jared-diener]

We have a use case to publish outcome data that identifies the top 3 industries where program completers are employed. Qdata only models numeric data, not categoric data such as these.

Adding a property such as "categoryObserved" to qdata:Observation, which would be a URI for a description of what was measured, would allow for this.

@philbarker FYI

[philbarker]

Yes, as Jared says, when we did the QData work, we only really considered numeric data, we didn't consider metrics/observations where the thing being measured is a category, e.g. the most common industry destination. So we can represent data for % of people who go into various industries (the is observation numeric data, represented as a percentage at points in a dimension of industries), but not the measurement of the Most Popular Industry (the observation is categoric data). An example with categoryObserved:

The range of categoryObserved would be rdfs:Resource (same as atPoint).
Definition would be something like "Class or type of thing observed." We might need a comment to explain that further.

[philbarker]

Example data here: https://docs.google.com/spreadsheets/d/1wEzoVlMKr56E4yhs-2dAWNhUp6HdJ80T_VULaT-naAk/edit?gid=1424232257#gid=1424232257 (contains data from data sets covering several quarters)

[philbarker]

Here are two options for showing the first example most popular destination from that spreadsheet: row 5, where the quarter2_naics1_code is 325412 Pharmaceutical Preparation Manufacturing

First if you want to explicitly state in a machine actionable way that the categories are Industries:

{
  "@type": "qdata:Observation",
  "qdata:isObservationOf": {
    "@type": "qdata:Metric",
    "ceterms:name": { "en": "Most Popular Industry" }
  },
  "qdata:observedCategory": {
    "@type": "ceterms:Industry",
    "ceterms:name": { "en": "Pharmaceutical Preparation Manufacturing" },
    "ceterms:industryType": {
        "@type": "ceterms:CredentialAlignmentObject",
        "ceterms:frameworkName": {"en": "NAICS"},
        "ceterms:codedNotation": "325412",
        "ceterms:targetName": { "en": "Pharmaceutical Preparation Manufacturing" }
    }
  }
}

(yeah, I know, you maybe wouldn't embed the bnodes; also you don't really need the name property on Industry and the targetName for the alignment object, you could just pick one)

Second, if you're OK to rely on the text of the metric and implicit understanding of what a NAICS code is:

{
  "@type": "qdata:Observation",
  "qdata:isObservationOf": {
    "@type": "qdata:Metric",
    "ceterms:name": { "en": "Most Popular Industry" }
  },
  "qdata:observedCategory": {
    "ceterms:industryType": {
      "@type": "ceterms:CredentialAlignmentObject",
      "ceterms:frameworkName": { "en": "NAICS" },
      "ceterms:codedNotation": "325412",
      "ceterms:targetName": { "en": "Pharmaceutical Preparation Manufacturing" }
    }
  }
}

BTW, to represent the whole dataset(s), I would suggest that the rows are various observations of various metrics (Number of exiters, Number of completers, number & rate employed etc.) made at various points along a dimension that has the Programs as points in it. There are other ways of doing it that may be simpler but would involve splitting it into more datasets, e.g. one for each program.

[jeannekitchens]

@philbarker @jared-diener I want to make sure we are thinking this through carefully and not treating a non-numeric outcome as a blank node.

The observed outcome may simply be a textual result (rdf:langString) with an optional alignment to a formal classification system when applicable. Not all observed outcomes are classifications, categories, or concepts in a controlled vocabulary.

For example:

Industries may align to NAICS classification
Occupations may align to SOC or ESCO classification
Jobs, employers, or transfer destinations are not classifications

Shorthand URI: qdata:observedOutcome
Label: Observed Outcome
Definition: Non-numeric outcome or result identified by an observation.
Usage Note: Use when the result of an observation is expressed as text rather than as a numeric value, such as the name of an industry, occupation, job, employer, transfer destination, or other outcome.
Domain: qdata:Observation
Range: rdf:langString

Shorthand URI: qdata:outcomeClassification
Label: Outcome Classification
Definition: Classification that identifies or further describes the observed outcome.
Usage note: Use when the observed outcome is associated with a formal classification or coding scheme, such as NAICS for industries or SOC for occupations.
Domain: qdata:Observation
Range: ceterms:CredentialAlignmentObject

{
"@type": "qdata:Observation",
"qdata:isObservationOf": {
"@type": "qdata:Metric",
"ceterms:name": { "en": "Primary Industry of Employment" }
},
"qdata:observedOutcome": {
"@value": "Pharmaceutical Preparation Manufacturing",
"@language": "en"
},
"qdata:outcomeClassification": {
"@type": "ceterms:CredentialAlignmentObject",
"ceterms:frameworkName": { "en": "NAICS" },
"ceterms:codedNotation": "325412",
"ceterms:targetName": { "en": "Pharmaceutical Preparation Manufacturing" }
}
}

Exmple without classification. Could also have an example that points to a published job.

{
"@type": "qdata:Observation",
"qdata:isObservationOf": {
"@type": "qdata:Metric",
"ceterms:name": { "en": "Primary Employment" }
},
"qdata:observedOutcome": {
"@value": "Junior Data Analyst",
"@language": "en"
}
}

[philbarker]

@jeannekitchens

Text descriptions of observations could be provided using the ceasn:comment property, or we could add ceterms:description; no need for observedOutcome with a text value but we could add it if you wish.

The range of categoryObserved proposed above is rdfs:Resource, so it could be an instance of anything that is not text, a blank or named node, either as a URI or an embedded object. All of what you describe above can be covered and more, and potentially with greater clarity about what type of thing is being observed and how it relates to other things, if that is desired.

[jeannekitchens]

@philbarker yes. We can/should reuse properties where it makes sense. https://credreg.net/qdata/terms/atPoint range is Resoource.

[philbarker]

@jeannekitchens

https://credreg.net/qdata/terms/atPoint range is Resource.

I know, but so what?

[jeannekitchens]

If the point of adding this property is to point to Credential Alignment Object, for USA, also need industryType, occupationType, intructionalProgramType an,d possibly rather than observationCategory, observationType.