Skip to content

Latest commit

 

History

History
92 lines (68 loc) · 3.94 KB

guidelines.md

File metadata and controls

92 lines (68 loc) · 3.94 KB

Guidelines

Introduction

  • aim of the document, intended audience, etc.
  • governance: who owns the document, how to contribute?

What are we talking about?

  • types of data to describe (cross-sectional, longitudinal...) with examples corresponding to the use cases
  • variables and questions
  • value domain representation
  • high-level description of the variable cascade (without reference to DDI)
  • what do we mean by repetitive and longitudinal contexts
  • what do we want to represent and why (provenance, reproducibility, machine-actionability...)

What does DDI-L provide?

DDI-Lifecycle provides a range of different structures which can used alone or in combination to manage item types. Item types structures which are important to the use cases include

  • Concepts
  • Conceptual Variables
  • Represented Variables
  • Variables (Instance variables)
    • code lists
      • code value
      • categories
    • numeric representation
    • substantive (valid) / sentinel (invalid) values
  • Questions
    • response domains
  • Universes
    • Population
      • Spatial
      • Temporal
  • Unit type
  • Measurement

Controlled Vocabularies

Controlled vocabularies (standardised and organised arrangements of words and phrases and provide a consistent way to describe data) though not specific to DDI-L they can be used to describe the usage of a specific item type. An example is type of instrument used in data collection: https://vocabularies.cessda.eu/vocabulary/TypeOfInstrument?lang=en

Schemes

These are for organising a single item type, such as questions, variables, universes and code lists to support management and reuse across an organistion.

Groups

These structures support discovery and organisation of item types by

  • referencing a specific item type within one or many schemes, e.g. variables related to Age
    • CLOSER Topics Level 1 and 2
    • Insee example from LFS
  • group the same item types which are not managed in a scheme e.g. waves of data collection in an ongoing study
    • CLOSER

Relationships

These support the definition of the location and type of a object being referenced for example variable representation can be described as code list and a reference to the target URN of the object code list. A code list can be referenced from many variables supporting its reuse.

BasedOn supports annotation of the relationship between a specific item type to the same item type, with a textual description and / or controlled vocabulary that describes the nature of the relationship. This supports the provenance of a specific item type to the same item type. e.g. variable B is based on Variable A.

Variable Cascade

Supports the specialisation of variables with a reference Conceptual -> Represented -> (Instance) Variable Note inherited item types are in bold)

  • Conceptual variable (concept + unit type)
  • Represented variable (concept + unit type + value representation (substantive values))
  • (Instance) Variable (concept + unit type + variable representation (substantive + sentinal values), universe reference, measurement, question reference etc)

Use cases

For each use case

  • description
  • what do we want to formalize and why?
  • precise example in pseudo-DDI
  • alternative possibilities with pros and cons for each

(for CLOSER use case, mention automated methods for finding relations: NLP...)

Links

What do we need (more than DDI-L)?

  • baseOn categories (cf. NACDA)

Conclusions

  • DDI provides different tools, additional ones may be needed, some are in CDI
  • There is no perfect answer, it depends on multiple factors (effort in metadata modelling, needs of users, etc.)