Module 3: Data types & formats, data distribution and re-use
Topic 1: Data types and formats
*Technical information
1. All data in computers based on digital electronics is represented as bits (alternating
between 0 and 1) at the most basic level.
2. The smallest addressable unit of data is usually a group of bits called
a byte (typically an octet of 8 bits).
3. The unit processed by machine code instructions is called a word (typically 32 or
64 bits).
Endianness refers to the order in which a computer arranges and reads bytes of
data in memory or during communication. There are two main types:
1. Big-endian (BE): the most significant byte is stored first. A system using the
BE format would store '0000 0001' exactly as it is.
2. Little-endian (LE): The order is reversed, so the least significant part comes
first. A little-endian system would store the same number as '0001 0000'
instead.
Endianness is important because different computer systems use different formats.
When data is shared between systems, the endianness must be the same, otherwise
the data could be misinterpreted.
a. Primitive Data Types
Primitive data types are the simplest form of data used by computers. These include
numbers, text and logical values.
Data Type Definition Example
Boolean Two possible values: A website asks if you agree to the
True and False terms and conditions. The system
stores your answer as True if you
accept, and False if you decline.
Integer Whole numbers The number of students enrolled in a
course is 250.
Floating-point Numbers with decimal A weather app displays the
numbers points temperature as 30.5°C.
Characters Single Letter, Number A password field requires at least one
or Symbol special character, such as @ or #.
String Text When you enter a name (“Nguyen
Minh An”) on a website, it is stored as
a string since it consists of multiple
characters.
b. Complex Data Types
While primitive data types store individual pieces of information, complex data types
help to organise and structure related data.
Data Definition Example
Type
List Ordered A grocery shopping list in an app
collection of (Apples, Milk, Bread). The list stores
items items in order, allowing users to check
them off one by one.
Set Unordered A university database keeps track of the
collection of different majors students are enrolled in.
unique items Since no two majors should be identical,
a set is used to prevent duplicates.
Recor Multiple A student database recording name, age,
d pieces of student ID and major.
information
about a
single entity
Table Organised This is a table!
records in
rows and
columns
Matrix Two- A digital image is stored as a matrix,
dimensional where each number represents a pixel’s
array of color or brightness.
numbers
arranged in
rows and
columns
Tensor Multi- Similar to storing a digital image in a
dimensional matrix, 3D models are stored as a tensor.
matrices
Graph Information Facebook’s Friend Network stores each
is displayed person as a node, connecting lines mean
through that they are friends. This structure helps
connected the system suggest people you may
nodes to know.
display
relationships
Tree Hierarchical A file system on your computer. The
system to main folder holds subfolders that contain
structure files, to help organise them efficiently.
information
BLOB Large binary Uploading a profile picture on social
data objects, media. The picture is stored as a BLOB
such as as it contains binary data (pixels).
photo or
video files.
CLOB A data type A wikipedia article is stored in a CLOB
used to store format since it contains a large amount
large of text that needs to be saved in a
amounts of structured way.
text in
databases.
c. Higher-level Data Types
Higher-level data types are often designed for specific applications, such as handling
images, videos or geographical information.
Data type Definition Example
Pixel-Based Formats Stores images as a grid of tiny A photograph taken on a
(JPEG or PNG) colored dots (pixels). Each smartphone is stored as a JPEG
pixel contains color file. If you zoom in too much,
information. the picture becomes blurry.
Vector-Based Represents images using A company logo stored as an
Formats (SVG) mathematical shapes, SVG file can be resized
allowing for infinite scaling infinitely without losing
without using quality. quality. This is why company
logos on websites look sharp on
all devices.
Video Codec and Defines how video files are A TikTok video is stored in an
Format compressed and stored, MP4 format using the H.264
balancing file size and codec to ensure it plays
playback quality for smoothly while keeping the file
streaming and storage. size small.
Geospatial Data Stores geographic information Google Maps stores location
like coordinates, maps and data in GeoJSON format,
boundaries. Used in which contains
navigation systems and digital latitude/longitude coordinates
mapping. and road networks.
c. Metadata
- It provides key details such as what the data represents, where it comes from and
how it should be used.
- For instance, a photo’s metadata may include the date, camera model and location,
whereas research data could contain information on collection methods and usage
limitations.
Topic 2: Contextuality of Data
o Data does not exist in isolation - it is shaped by where, when and why it was
created.
o To use data ethically and accurately, we must always ask about its original
context before drawing conclusions.
o Context helps us understand people beyond numbers.
o Data does not explain itself, we must interpret it carefully.
o Each dataset only tells part of the story.
o Data is never neutral - it reflects who collected it and how it was used.
o Overlapping contexts shape data differently for residents, care workers and nursing
home operators.
o External users must be cautious when interpreting data without understanding its
origins.
o More transparency in metadata is needed to improve data interpretation.
Understanding context is essential for making fair, ethical and meaningful use
of data.
Topic 3: Purpose-dependency of data
Assumption abt causal relationships
Eg: using data on financial detection that was collected to detect fraud for credit risk
assessments. Fraud detection is concerned with spotting anomalies that may indicate
criminal activity, while credit risk assessment evaluates financial stability and
repayment ability over time. Repurposing fraud detection models for credit
assessment leads to misclassifications of behaviours and transactions, and
consequently to false results in credit risk assessments, with severe implications for
credit applicants.
Selection of metrics:
Data collected for one purpose cannot always be meaningfully repurposed without
adjusting for its original context, measurement scales, and intended use. Ignoring
these factors can lead to incorrect conclusions, flawed decisions, or even nonsense
results.
Harmful feedback loops:
A harmful feedback loop with a situation where data was originally collected for one
purpose, but later, a second purpose was added. As a result, the people being observed
change their behavior to meet the second purpose, which makes it impossible to
achieve the original goal.
Scales of Measurement
Different scales of measurement clarify and quantify data in various ways, depending
on the properties of the values involved. The four main types are nominal, ordinal,
interval and ratio. The level of information provided by each scale determines the
types of mathematical operations that can be performed.
The nominal scale categorises data into distinct groups, which have no inherent order
or relationship between them. It is used for classification purposes only. Examples
include types of animal (e.g. cat, dog, rabbit), gender, nationality or language.
Although numbers can be assigned to categories (e.g. 1 for 'cat', 2 for 'dog', 3 for
'rabbit'), these numbers do not have any mathematical significance.
The ordinal type allows data to be ranked or ordered, but does not specify the exact
differences between values. It is commonly used in rating systems, such as survey
responses (e.g. 'disagree', 'somewhat agree', 'agree') or competition rankings (1st, 2nd,
3rd). As ordinal values have no absolute value, they only allow ranking in terms of
highest to lowest. For example, the difference between ranks 1 and 2 may be greater
than the difference between ranks 7 and 8; ordinal values only indicate the order of
the ranking.
Although the interval scale measures differences between values and provides
meaningful distances, it lacks a true zero point. This means that meaningful ratios
cannot be calculated. Eg: 0°C does not indicate an absence of temperature; rather, 0°C
was chosen as the point at which water freezes. Other examples include dates
(measured from an arbitrary starting point) and geographic coordinates.
The ratio scale has all the properties of the interval scale, but also includes a
meaningful zero point. This allows comparisons to be made using multiplication and
division. Eg: mass, length, time, and energy
Topic 4: Temporality of Data
The right to be forgoten: a legal and ethical concept that allows individuals to request
the removal of personal data from digital systems, particularly when it is outdated,
irrelevant, or harmful.
Drifts in meaning overtime. Eg: The word “gay” arrived in English during the 12th
century from Old French “gai,“ with its primary meaning being for
centuries “joyful,” “carefree,” and “bright and showy.” It was not until the 20th
century that the word began to be used to mean specifically “homosexual.”
PredictionS and Forcasting: Forecasting models often assume temporal
stability, but when social, economic, technological, or environmental factors
change, past data can become unreliable. For example, a machine learning model
predicting energy consumption based on past trends may break down when faced with
sudden technological shifts or regulatory changes.