Introduction
Welcome to our final lesson in the "Structuring Data with SAD Metadata" module. In our previous lessons, we explored each component of the SAD mnemonic individually. We learned that Structural metadata provides the data's blueprint, Administrative metadata is the rulebook for its management, and Descriptive metadata acts as the public-facing label that tells us what the data is about.
Today's lesson synthesizes this knowledge. Our goal is to classify a given piece of metadata as structural, administrative, or descriptive using examples from bra design and relationship research. This is a crucial skill for any data practitioner, as it allows you to quickly understand how a dataset is organized, how it should be handled, and what it contains. By the end of this lesson, you will be able to look at a piece of information about a dataset and confidently categorize it, solidifying your understanding of the SAD framework.
1. A Quick Review: Structural, Administrative, and Descriptive
Before we begin classifying, let's briefly refresh our memory on the core function of each metadata type. A single dataset will have all three types working together.
- Structural Metadata (S): The blueprint. It describes how the data is organized and put together. It's about the data's format, structure, and relationships.
- Core Question: How is this data constructed and organized?
- Administrative Metadata (A): The logbook or rulebook. It's used to manage the data throughout its lifecycle. It deals with technical management, access rights, and preservation.
- Core Question: How is this data managed, accessed, and preserved?
- Descriptive Metadata (D): The label or summary. It describes the intellectual content of the data to enable discovery and understanding.
- Core Question: What is this data about?
For a concise summary of these definitions, let's turn to a resource we've touched on before.
Types of Metadata in Data Governance: A Complete Guide - OvalEdge
This article from OvalEdge, 'Types of Metadata in Data Governance', has a helpful FAQ section that neatly summarizes the three types.
Please read the first question in the 'FAQs' section, titled 'What are the three main types of metadata?'. This provides a clear, high-level definition for each of the three types we will be working with today.
With these definitions refreshed, we can now move on to the practical task of classification.
2. Classification Exercise: A Bra Fit Dataset
Imagine we are working with a dataset similar to the "Rent the Runway" dataset, which contains rich information about clothing fit. Let's call our version the "Bra Fit and Style Dataset." It contains data from thousands of customers who have purchased bras and provided feedback.
The article "RentTheRunway Dataset: Deep Dive into Fashion Fit..." provides excellent real-world examples of the kinds of data points we would encounter.
RentTheRunway Dataset: Deep Dive into Fashion Fit, Context, and ...
To ground our exercise in a realistic context, let's look at the data structure of a real-world fashion dataset. This will provide the items we will classify.
First, quickly scan the bullet points under the 'Key Features & Data Structure of the RTR Dataset' section to familiarize yourself with the data fields (e.g., 'user_id', 'bust size', 'timestamp'). Then, look at the YAML file under the heading 'rtr_model_schema.yaml'. This shows how these data points are formally structured, which is itself a form of metadata.
Now, let's classify the following metadata elements from our hypothetical "Bra Fit and Style Dataset". For each item, decide if its primary function is Structural, Administrative, or Descriptive.
Item 1: The definition of the user_id column as the "primary key" in the customer table, which uniquely identifies each customer.
Your classification:
**Structural**. The concept of a primary key defines the structure and relationships within the database. It specifies that this column is the unique identifier used to link to other tables, which is purely about the data's organization.Item 2: A policy stating, "All personally identifiable information, including user measurements, must be deleted from the active database 5 years after the customer's last purchase."
Your classification:
**Administrative**. This is a classic example of administrative metadata. It's a rule that governs the data's lifecycle (a retention period) and is crucial for legal compliance and data management. It doesn't describe what the data *is* or how it's formatted, but rather how it's *handled*.Item 3: The text "34B" in the bust_size column for a specific customer.
Your classification:
**Descriptive**. This value describes a specific attribute of the customer. The metadata is the field name `bust_size` and its associated description: "The self-reported bust size of the customer." This element helps us understand the content and context of the data for a particular record.Item 4: The data type for the review_text column is defined as TEXT.
Your classification:
**Structural**. The data type specifies the technical format of the data stored in that column. It's part of the database schema—the blueprint—that dictates how data is physically stored and organized.Item 5: A summary attached to the dataset: "Contains 10,000 anonymized customer records, including fit feedback, style preference, and purchase history for lingerie items from 2020-2023."
Your classification:
**Descriptive**. This is a perfect example of an abstract or description. Its sole purpose is to tell a potential user what the dataset contains, making it discoverable and understandable without needing to open the file.Item 6: A record in a log file showing that analyst_jane_doe accessed the dataset on 2024-03-15T10:00:00Z.
Your classification:
**Administrative**. This is access-control and usage metadata. It's used for security audits and managing who has rights to the data. It's about the *management* of the data resource.3. Classification Exercise: A Polyamorous Relationship Study
Now, let's apply the same logic to a dataset from your other area of expertise. Imagine a dataset from a longitudinal study on communication and satisfaction within single-female-multiple-male polyamorous relationships (V-triads).
Here are several metadata elements related to this study. Classify each one as Structural, Administrative, or Descriptive.
Item 1: A document specifying the relationships between three data tables: Participants, Dyads, and Weekly_Checkins. The participant_id in the Participants table links to two fields in the Dyads table.
Your classification:
**Structural**. This describes the relational schema of the database. Defining tables and the links between them is the essence of structural metadata. It is the blueprint of how the data points are organized and interconnected.Item 2: The keywords section of the dataset's entry in a research data repository reads: polyamory, non-monogamy, relationship satisfaction, communication, V-triad.
Your classification:
**Descriptive**. These keywords describe the intellectual content of the dataset. Their purpose is to make the data discoverable to other researchers searching for these topics.Item 3: The file format of the interview transcripts is .docx, and the survey data is in .csv.
Your classification:
**Administrative**. While it feels technical, the file format is primarily used for management purposes. It dictates the software needed to open the file (preservation and accessibility) and is a key piece of technical information for managing the data asset. The *content* of the file is what's described by descriptive metadata; the *container* is handled by administrative metadata.Item 4: The informed consent form, which states, "Your data will be de-identified, stored on an encrypted server, and will only be accessible to the primary research team."
Your classification:
**Administrative**. This metadata defines the access rights and security measures for the data. It's part of the governance framework managing the resource and ensuring ethical and legal compliance. It doesn't describe the content, but the rules surrounding it.Item 5: In the dataset's codebook, the variable satisfaction_score is defined as: "An integer from 1-7, representing the participant's self-reported relationship satisfaction, collected via the aformentioned Weekly_Checkins survey."
Your classification:
**Descriptive**. This is a variable-level description. It explains what the data in the `satisfaction_score` column actually means. Without this description, an analyst might not know the scale or context of the number, making it difficult to interpret. The variable name `satisfaction_score` and its data type (`integer`) are structural, but the explanation of what it *represents* is descriptive.Conclusion
In this lesson, we put your knowledge of the SAD framework into practice. We moved from defining the three types of metadata to actively classifying them in scenarios relevant to your work. This exercise highlights how Structural, Administrative, and Descriptive metadata are not isolated concepts but a cohesive system for making data organized, manageable, and understandable.
Key Takeaways:
- Structural metadata is the blueprint (e.g., data types, primary keys, table relationships).
- Administrative metadata is the rulebook and logbook (e.g., access rights, retention policies, file formats).
- Descriptive metadata is the summary label (e.g., titles, keywords, abstracts, codebook definitions).
- The same dataset will always contain all three types, and correctly classifying them is key to effective data handling.
Preview of the Next Lesson:
We have now completed our module on how to describe and structure data using the SAD framework. You are now equipped to document a dataset thoroughly.
Next, we will shift our focus from the data itself to the claims we make based on it. We will begin Module 5: Validating Claims with FLOATER. Our first lesson will address a fundamental question in statistics and research: why must claims derived from data be systematically validated before being accepted? This will set the stage for learning the FLOATER framework as a tool for critical evaluation.