Quantitative data (量的データ) expresses a measured size in numbers, so adding and averaging its values gives meaningful results.
It is recorded as numbers with units, such as height, sales amounts, or store visitor counts. Totals, averages, and standard deviations apply as they are, and histograms or scatter plots show how the values are spread and how they relate.
AnalogyThink of the readings on a bathroom scale. With 60 kg and 70 kg, "the average is 65 kg" and "the difference is 10 kg" are results that mean something as they stand.
How to rememberIn English it is "quantitative," from the same root as "quantity": data that answers "how much."
Qualitative data (質的データ) marks kinds or categories, so adding or averaging it means nothing.
It records which category something falls into, such as blood type, occupation, or a satisfaction level (dissatisfied, neutral, satisfied). Summaries count the cases or shares in each category, and pie charts or 100% stacked bar charts often show the breakdown.
AnalogyThink of splitting a school sports day into red and white teams. Adding "red" to "white" produces nothing; all you can count is how many people are on each team.
How to rememberIn English it is "qualitative," from the same root as "quality": data that answers "what kind."
Exam points
Mix-upData written in digits can still be qualitative data (質的データ). Membership numbers and postal codes are labels for categories, so adding or averaging them means nothing.
Primary data (1次データ) is data you collect yourself, fresh, for the question you are trying to answer now.
It comes from your own surveys, interviews, in-store observation, experiments, and similar methods. Because you design the questions and choose the subjects, it tends to fit the purpose well, but collecting it takes time and money.
AnalogyLike vegetables you pick from the field yourself. You choose exactly the variety you want, but it takes effort.
How to remember"Primary" means obtained first-hand, the same idea as "first-hand information."
Secondary data (2次データ) is data already collected for another purpose that you reuse in your own analysis.
Examples include national and local government statistics, industry association reports, and materials published by research firms. It is quick and cheap to obtain, but the survey period or how items are defined may not match your purpose, so check how it was produced before relying on it.
AnalogyLike a ready-made dish bought at the store. You can eat it right away, but you do not get to choose the seasoning.
Exam points
Mix-upRecords already held inside the company also count as secondary data (2次データ) when they were gathered for some other purpose. What decides it is not whether they came from inside or outside the company, but whether they were newly collected for this purpose.
Structured data (構造化データ) is data organized into a table of rows and columns according to fields (columns) decided in advance.
Typical examples are tables in a relational database and a sales list in a spreadsheet. Because each value belongs to a known field, a computer can search by condition and total or sort by field without any extra preparation.
AnalogyLike an application form with printed boxes. Names go in the name box and dates in the date box, so even hundreds of forms can be tallied the same way.
Unstructured data (非構造化データ) is text, images, audio, video, and other data that do not fit into fixed fields.
Examples include email body text, SNS posts, security camera footage, and recorded phone calls. It cannot be searched or totaled like a table as it stands, so features are first extracted, for instance by picking out words from text or by image recognition, before the data is analyzed.
AnalogyLike the free-comment box on a survey. Each person writes something different at a different length, so you cannot count anything until you read it and pull out the content.
Exam points
Mix-upFile format does not decide it. Among plain text files, a CSV file whose fields are fixed by delimiters is structured data (構造化データ), while free-form prose such as meeting minutes is unstructured data (非構造化データ).
Metadata (メタデータ) describes not the contents of data but its attributes, such as when it was created, who created it, and its format.
Category
Photo file example
Data itself
The image that was captured
Metadata
Date and time taken
Metadata
Camera model and location
Metadata
File format and size
It is also called "data about data." Without opening the data itself, you can use metadata to find, classify, and manage the data you need. In a database, the definitions that set each field's name, type, and unit are also metadata.
AnalogyLike the classification label on a library book. You can find it on the shelf by title, author, and call number without reading a page.
How to remember"Meta" comes from Greek and means "beyond" or "one level up." Metadata sits one level above the data and describes it.
Time-series data (時系列データ) records the same subject again and again as time passes.
Daily temperatures, monthly sales, and yearly population counts are examples. Lined up in time order, they reveal trends up or down and seasonal swings, and line charts are the usual way to show how they change.
AnalogyLike height marks carved on a doorframe. The same child is measured each year, so you can see how they grew.
How to remember"Series" means things lined up one after another, so a time series is simply values lined up in time order.
Exam points
ExamA data warehouse (データウェアハウス) is a collection of data organized by subject and accumulated over time to support management decisions.
Cross-section data (クロスセクションデータ) takes a slice at one point in time and gathers many subjects side by side.
Store (sales, 10,000 yen)
Apr
May
Jun
North
120
135
150
South
90
95
80
Station
200
190
210
Examples include each store's sales in the same month or each region's population in a given year, and it suits comparing subjects with one another. In the table above, reading one column (a month) down gives cross-section data, and reading one row (a store) across gives time-series data.
AnalogyLike a class photo taken one morning. You can compare everyone at that moment, but nothing about how they change afterward appears in it.
How to remember"Cross section" means a slice. Picture one cut taken straight across the flow of time.
Exam points
Mix-upIt is not the same as a cross tabulation (クロス集計表). A cross tabulation is a table that totals data by combining two fields, while this term describes when and about what the data was collected, a different angle altogether.
Time-series data vs. Cross-section data
Aspect
Time-series data
Cross-section data
Key differenceWhat is held fixed
The subject
The point in time
How values line up
In time order
By subject
Suited to
Changes and trends
Comparing subjects
Key points
The kind of data decides which summaries you can use, so before analyzing, work out which category your data belongs to.
The ways of classifying are independent of one another. For example, yearly population figures taken from government publications are quantitative, secondary, and recorded over time all at once.
Metadata (メタデータ) is supporting information that describes the data itself, and it is the key to searching, managing, and reusing large amounts of data.