Week 1: Types of Data & Scales of Measurement

3787 words

19 min read

2026-09-12T00:00:00.000Z

Visual companion

Python

Type and operator map

Python Week 1: the first filter for runtime behavior

Open gallery Related surface

View

Revision summary

What this note is really saying

Short form

Data classification, categorical vs numerical variables, scales of measurement (nominal, ordinal, interval, ratio), and observational studies. # Statistics for Data Science I — Week 1: Types of Data & Scales of Measurement IITM’s Week 1 introduces **types of data, descriptive and inferential statistics, and scales o...

Statistics for Data Science I — Week 1: Types of Data & Scales of Measurement

IITM’s Week 1 introduces types of data, descriptive and inferential statistics, and scales of measurement.

TL;DR

Statistics starts with one question:

What kind of data do I have, and what can I legitimately do with it?

The central chain is:

\text{Population} \rightarrow \text{Sample} \rightarrow \text{Data} \rightarrow \text{Describe} \rightarrow \text{Infer}

The highest-value ideas are:

population vs sample
parameter vs statistic
descriptive vs inferential statistics
categorical vs numerical data
discrete vs continuous data
nominal / ordinal / interval / ratio scales
why the measurement scale controls valid mathematical operations

1. What is statistics?

Statistics is the discipline of:

collecting data
organizing data
summarizing data
analyzing data
drawing conclusions from data

It is not merely "calculating averages."

At a high level:

\boxed{ \text{Statistics}= \text{learning about variation and uncertainty from data} }

2. What is data?

Data are recorded observations about entities.

Examples:

a person's age
exam score
city
height
product rating
number of purchases
blood group
temperature

A single measured characteristic is usually called a variable.

3. Observation vs variable

Suppose we have:

Student	Age	Score	City
A	19	82	Rajkot
B	21	75	Ahmedabad
C	20	91	Junagadh

Each row is an observation.

Each column is a variable.

So:


text
Observation = one entity/record
Variable = one characteristic measured across observations

4. Population

The population is the complete collection of individuals or objects we want to study.

Examples:

all IITM BS students
all voters in India
all light bulbs produced by a factory this month
every transaction made by a company in 2026

Population does not necessarily mean "people."

It means the complete target collection.

5. Sample

A sample is a subset of the population that is actually observed.

If:

P=\text{all IITM BS students}

and we survey:

500

students, those 500 form the sample.

So:

\boxed{ \text{Sample}\subseteq\text{Population} }

6. Why do we use samples?

Often studying the whole population is:

too expensive
too slow
impossible
destructive
unnecessary

Example:

If a factory wants to test the breaking strength of every bulb by destroying it, testing the entire population would leave nothing to sell.

So we sample.

7. Parameter

A parameter is a numerical characteristic of a population.

Examples:

true population mean
true population proportion
true population variance

Common notation:

Population mean:

\mu

Population variance:

\sigma^2

Population standard deviation:

\sigma

Population proportion:

p

8. Statistic

A statistic is a numerical characteristic calculated from a sample.

Examples:

Sample mean:

\bar{x}

Sample variance:

s^2

Sample standard deviation:

s

Sample proportion:

\hat p

9. Parameter vs statistic

This distinction is fundamental.

Population	Sample
parameter	statistic
$\mu$	$\bar{x}$
$\sigma$	$s$
$p$	$\hat p$

A statistic is often used to estimate a parameter.

For example:

\bar{x} \approx \mu.

10. Example

Suppose all 10,000 students have an actual average height:

\mu=169.8\text{ cm}.

You sample

200

students and calculate:

\bar{x}=170.4\text{ cm}.

Then:

$169.8$ cm is a parameter
$170.4$ cm is a statistic

11. Descriptive statistics

Descriptive statistics summarizes observed data.

Typical tools include:

tables
charts
mean
median
mode
range
variance
standard deviation
percentages

It answers:

What does the data we actually observed look like?

12. Example of descriptive statistics

Suppose five scores are:

60,70,70,80,90.

We may report:

Mean:

74

Median:

70

Range:

90-60=30.

Those statements summarize those five observations.

No broader population conclusion is required.

13. Inferential statistics

Inferential statistics uses sample data to make statements about a wider population.

Examples:

estimating average income of a city from sampled households
predicting election support using a poll
determining whether a medicine works from a clinical trial
estimating defect rate using sampled products

Conceptually:

\boxed{ \text{Sample} \rightarrow \text{Inference about population} }

14. Descriptive vs inferential

Suppose we survey:

500

students.

Descriptive

62% of the surveyed students prefer online exams.

This describes the sample.

Inferential

We estimate that about 62% of all students prefer online exams.

Now we are generalizing beyond the observed sample.

15. Why inference contains uncertainty

A sample is not usually identical to the population.

Different samples may produce different values.

For example:

\bar{x}_1=71

\bar{x}_2=73

\bar{x}_3=70.

This variation is why probability becomes central later in Statistics 1.

Statistical inference does not normally say:

"The population value definitely equals this."

It quantifies uncertainty.

16. Variable

A variable is a characteristic that can take different values.

Examples:

age
gender category
annual salary
number of siblings
exam grade
temperature

The first major classification is:

\boxed{ \text{Categorical} \quad\text{vs}\quad \text{Numerical} }

17. Categorical data

Categorical data represents labels or groups.

Examples:

blood group
city
marital status
operating system
eye color
department

The values indicate category membership rather than meaningful arithmetic quantities.

18. Numerical data

Numerical data represents quantities where arithmetic usually has meaningful interpretation.

Examples:

height
weight
age
income
number of purchases
temperature

Numerical data is often further divided into:

\boxed{ \text{Discrete} \quad\text{and}\quad \text{Continuous} }

19. Discrete data

Discrete variables take separated, countable values.

Usually obtained by counting.

Examples:

number of children
number of cars
number of emails received
number of goals scored

Possible values might be:

0,1,2,3,\dots

You generally cannot have:

2.73

children.

20. Continuous data

Continuous variables can take values across an interval.

Usually obtained by measurement.

Examples:

height
weight
time
temperature
distance

A height may theoretically be:

170.1,\quad170.12,\quad170.123,\dots

depending on measurement precision.

21. Count vs measurement mental model

A useful shortcut:

Discrete

"How many?"

Continuous

"How much?"

Examples:

Number of phones:

\text{discrete}

Phone weight:

\text{continuous}

22. But be careful with decimal values

Having decimal notation does not automatically mean continuous.

Suppose a rating system allows only:

0,\ 0.5,\ 1,\ 1.5,\dots,5.

Even though decimals occur, values are from a fixed countable set.

So it is discrete.

The issue is not:

"Does it contain decimals?"

The issue is:

"Can it theoretically vary continuously?"

23. Identifier numbers are not numerical measurements

Suppose:


text
Student ID = 104293

The value is written using digits.

But arithmetic like:

104293+104294

has no meaningful statistical interpretation.

So Student ID is effectively categorical/nominal data.

This is a common trap:

Numbers used as labels are not necessarily quantitative variables.

24. Scales of measurement

Variables are also classified by measurement scale.

The classical four levels are:

\boxed{ \text{Nominal} \rightarrow \text{Ordinal} \rightarrow \text{Interval} \rightarrow \text{Ratio} }

Each level carries more mathematical structure.

25. Nominal scale

Nominal data consists of categories with no inherent ordering.

Examples:

blood group
country
eye color
operating system
department

For blood type:

A,B,AB,O

there is no meaningful statement such as:

A<B.

They are simply labels.

26. What can we do with nominal data?

Meaningful operations include:

equality
inequality
counting frequencies
percentages
mode

For example:

\text{Android}=\text{Android}

or:

\text{Android}\neq\text{iOS}.

But averaging category labels makes no sense.

27. Ordinal scale

Ordinal data has an inherent order.

Examples:

satisfaction: poor / fair / good / excellent
class rank
education level
star rating
competition placements

You can say:

\text{excellent}>\text{good}>\text{fair}.

But the distances between categories are not necessarily equal.

28. Ordinal example

Suppose satisfaction levels are encoded:

1=\text{poor}

2=\text{fair}

3=\text{good}

4=\text{excellent}.

Although numbers are used, we cannot automatically conclude:

the gap between poor and fair is exactly equal to the gap between good and excellent.

The numbers may just encode rank.

29. Nominal vs ordinal

Nominal:

categories only.

Ordinal:

categories + meaningful order.

Examples:

Variable	Scale
Blood type	nominal
City	nominal
Race position	ordinal
Satisfaction level	ordinal

30. Interval scale

Interval data has:

meaningful order
meaningful differences
equal intervals

But it lacks a meaningful absolute zero.

Classic example:

\text{temperature in Celsius}

Difference between:

20^\circ C

and:

30^\circ C

is the same size as the difference between:

30^\circ C

and:

40^\circ C.

Both differences are:

10^\circ C.

31. Why Celsius has no true zero

0^\circ C

does not mean "no temperature."

Therefore it is not meaningful to say:

40^\circ C

is "twice as hot" as:

20^\circ C.

The ratio:

\frac{40}{20}=2

does not have the physical meaning implied by a ratio scale.

32. Ratio scale

Ratio data has:

meaningful ordering
equal intervals
meaningful differences
a true zero

Examples:

height
weight
distance
age
duration
income
number of items

Because zero represents absence of the quantity, ratios are meaningful.

33. Example of meaningful ratio

If one object weighs:

20\text{ kg}

and another:

10\text{ kg},

then it is meaningful to say:

20=2\times10.

So the first weighs twice as much.

That makes weight ratio-scale data.

34. The four scales as increasing structure

Think:


text
Nominal
   ↓
ordering added
Ordinal
   ↓
equal differences added
Interval
   ↓
true zero added
Ratio

So:

\boxed{ \text{Nominal} < \text{Ordinal} < \text{Interval} < \text{Ratio} }

in terms of mathematical information available.

35. Master comparison table

Scale	Categories	Ordered	Equal intervals	True zero
Nominal	✓	✗	✗	✗
Ordinal	✓	✓	✗	✗
Interval	✓	✓	✓	✗
Ratio	✓	✓	✓	✓

This table is worth memorizing.

36. Examples by scale

Nominal

religion category
eye color
postal code
phone number
city

Ordinal

movie rating
class rank
satisfaction category
education level

Interval

Celsius temperature
Fahrenheit temperature
calendar year in ordinary interpretation

Ratio

Kelvin temperature
height
weight
time duration
age
income

37. Kelvin vs Celsius

Kelvin has a meaningful zero:

0K

corresponds to absolute zero.

So Kelvin temperature is treated as ratio scale.

Celsius has an arbitrary zero point:

0^\circ C

so Celsius is interval scale.

This contrast is a standard exam example.

38. Calendar years

Years such as:

2000,\quad2010,\quad2020

have meaningful equal differences:

2010-2000=10

2020-2010=10.

But year zero is not a "complete absence of time."

So calendar year is typically treated as interval-scale.

39. Age

Age has a meaningful zero:

0\text{ years}

means zero elapsed age from birth.

Differences and ratios are meaningful in the ordinary measurement sense.

Thus age is typically:

\boxed{\text{ratio}}

40. Ordinal numeric encoding trap

Suppose:

1=\text{bad}

2=\text{okay}

3=\text{good}.

Although:

3-2=1

and:

2-1=1,

that numerical equality does not prove equal psychological distance between categories.

The numerical encoding does not upgrade ordinal data to interval data.

41. Binary categorical data

Variables with two categories are categorical.

Examples:

yes/no
pass/fail
present/absent

Even if encoded:

0,1,

the underlying variable may still be nominal.

Coding does not determine the scale.

42. Qualitative vs quantitative

Another common terminology:

Qualitative

Usually categorical.

Quantitative

Usually numerical.

So:

\boxed{ \text{Categorical}\approx\text{Qualitative} }

and:

\boxed{ \text{Numerical}\approx\text{Quantitative} }

depending on textbook terminology.

43. Cross-sectional data

Cross-sectional data records many entities at approximately the same point in time.

Example:

salaries of 1,000 employees in September 2026.

The variation is primarily across individuals.

44. Time-series data

Time-series data records a variable across time.

Example:

India's monthly inflation rate from 2020–2026.

The ordering in time matters.

45. Panel / longitudinal intuition

If you repeatedly observe the same entities over time, you get longitudinal or panel-style data.

Example:

monthly spending of the same 100 households for three years.

This combines:

cross-sectional variation
temporal variation

This terminology may appear later even if Week 1 keeps things simpler.

46. Raw data vs processed information

Suppose raw values are:

65,72,84,91,68.

A summary such as:

\bar{x}=76

compresses those observations.

Statistics frequently turns:

\boxed{ \text{raw data} \rightarrow \text{useful summaries} }

But every summary discards some information.

That tradeoff matters.

47. Census

A census collects information from every population member.

If every household in a country is surveyed, that's a census.

Advantage:

potentially complete coverage

Disadvantages:

costly
slow
administratively difficult
still vulnerable to measurement/nonresponse errors

A census is not automatically error-free.

48. Sample survey

A sample survey collects information only from a subset.

If appropriately selected, the sample can provide useful inference about the population at much lower cost.

The quality depends heavily on how the sample is selected, not merely its size.

Sampling methods come later.

49. Representative sample

A representative sample resembles the relevant structure of the population sufficiently for the intended inference.

A huge biased sample can still be bad.

Example:

Suppose you want opinions of all citizens but survey only visitors to one specialized website.

Even:

1,000,000

responses may be systematically biased.

So:

\boxed{ \text{Large sample}\neq\text{automatically representative} }

50. Bias

Bias is a systematic tendency for a method to produce distorted results.

Possible sources include:

sampling bias
nonresponse
poorly worded questions
measurement errors
selection effects

Random variation and bias are different problems.

51. Observational units

An observational unit is the basic entity on which variables are measured.

Examples:

Dataset of students:

\text{unit}=\text{student}

Dataset of countries:

\text{unit}=\text{country}

Dataset of transactions:

\text{unit}=\text{transaction}

Knowing the observational unit prevents confusion about what rows actually represent.

52. Example classification

Consider:

Number of messages sent by a user per day.

This is:

numerical
discrete
ratio scale

Why?

It's a count:

0,1,2,\dots

and:

0

means no messages.

53. Example classification

Time needed to finish an exam.

This is:

numerical
continuous
ratio scale

A duration of:

0

means no elapsed time.

54. Example classification

User's preferred operating system.

This is:

categorical
nominal

There is no inherent ordering:

\text{Linux},\text{Windows},\text{macOS}

from the variable itself.

55. Example classification

Finishing position in a race.

Values:

1^\text{st},2^\text{nd},3^\text{rd},\dots

This is ordinal.

Rank tells order but not equal performance gaps.

The time gap between first and second may be:

0.1\text{s},

while second to third may be:

20\text{s}.

56. Example classification

Temperature in Celsius.

numerical
usually continuous
interval scale

57. Example classification

Height.

numerical
continuous
ratio scale

58. Example classification

Number printed on a cricket jersey.

Although numerical-looking:

18,\quad45,\quad7

the numbers serve primarily as labels.

So this variable is nominal.

59. Why scale matters

Measurement scale determines which operations make sense.

Suppose categories are encoded:

1=\text{red},\quad 2=\text{blue},\quad 3=\text{green}.

Then computing:

\frac{1+2+3}{3}=2

and concluding "average color = blue" is meaningless.

Numbers do not automatically legitimize arithmetic.

60. A useful operation hierarchy

Roughly:

Nominal

Can ask:

=,\neq

Ordinal

Can additionally ask:

<,>

Interval

Can additionally meaningfully compare differences:

x-y

Ratio

Can additionally meaningfully compare ratios:

\frac xy

That is a strong mental model for the four scales.

Common mistake: sample = population

No.

Population is the full target group.

Sample is the observed subset.

Unless a census is performed:

\text{sample}\neq\text{population}.

Common mistake: statistic = parameter

No.

Statistic:

\text{sample}

Parameter:

\text{population}

A statistic may estimate a parameter.

Common mistake: descriptive means "simple"

Descriptive vs inferential is not about difficulty.

A sophisticated visualization can still be descriptive.

A simple sample proportion used to generalize to a population becomes inferential.

The distinction is about scope of the conclusion.

Common mistake: all numbers are numerical data

Phone numbers, IDs, jersey numbers and PIN codes use digits but often represent nominal categories/identifiers.

Ask:

Does arithmetic on these values mean anything?

If no, treat them as labels.

Common mistake: discrete = integer

Often, but not necessarily.

A discrete variable can take noninteger values if those possibilities form a countable separated set.

Example:

0,\ 0.5,\ 1,\ 1.5,\dots

can still be discrete.

Common mistake: continuous means infinitely precise measurement

Actual measuring instruments always have finite precision.

A person's measured height may be recorded as:

178\text{ cm}.

The underlying variable is still conceptually continuous.

Variable type concerns the underlying possible values, not just displayed precision.

Common mistake: ordinal data has equal spacing

No.

Ordinal tells ordering only.

If:

A>B>C,

you know rank.

You do not necessarily know:

A-B=B-C.

Common mistake: zero always means ratio scale

Not every dataset containing the numeric value zero has a true zero.

Celsius contains:

0^\circ C

but zero does not mean absence of temperature.

"True zero" has a conceptual meaning.

69. Classification workflow

When given a variable, ask:


text
1. Is it a category or a quantity?
   ├── category → categorical
   └── quantity → numerical

2. If numerical:
   ├── countable separated values → discrete
   └── continuum → continuous

3. Measurement scale:
   ├── labels only → nominal
   ├── meaningful order → ordinal
   ├── equal intervals, arbitrary zero → interval
   └── equal intervals + true zero → ratio

This handles most Week 1 classification questions.

70. Complete example

Suppose a university records:

Variable	Example
Student ID	29183
Degree year	1st
Exam marks	86
Height	176.4 cm
City	Junagadh

Classification:

Student ID

Categorical, nominal.

Degree year

Often ordinal because year 1 < year 2 < year 3 in progression.

Exam marks

Numerical; typically discrete depending on scoring system; ratio/interval interpretation can depend on context, so don't automatically overclaim ratios.

Height

Numerical, continuous, ratio.

City

Categorical, nominal.

The context of a variable can matter.

71. Data classification is not always purely mechanical

Some variables can be interpreted differently depending on how they are defined.

Example:

"Age"

Exact age measured to arbitrary precision:

20.538\text{ years}

→ continuous.

Age recorded only as completed years:

20

→ discrete recorded data.

So always read the wording of the question carefully.

72. Week 1 conceptual map


text
STATISTICS
│
├── Population
│   └── parameter
│
├── Sample
│   └── statistic
│
├── Descriptive
│   └── summarize observed data
│
└── Inferential
    └── learn about population
          from sample

DATA
│
├── Categorical
│   ├── nominal
│   └── ordinal
│
└── Numerical
    ├── discrete
    └── continuous
         │
         ├── interval
         └── ratio

73. What you should actually master

By the end of Statistics 1 Week 1, you should be able to:

Define population and sample.
Distinguish parameter from statistic.
Distinguish descriptive from inferential statistics.
Identify observational units and variables.
Classify variables as categorical or numerical.
Distinguish discrete from continuous data.
Recognize numeric labels as categorical when appropriate.
Distinguish nominal from ordinal scales.
Distinguish interval from ratio scales.
Explain why Celsius is interval but Kelvin is ratio.
Explain why numeric coding does not change a variable's actual measurement scale.
Classify realistic variables from context.

If those are comfortable, Stats 1 Week 1 is under control.

Quick Recall & Flashcards

Test your active recall before proceeding to practice problems. Format:

Question :: Answer.

Week 1 Diagnostic: Which scale of measurement applies to 'temperature measured in Celsius'? :: Answer: C — Celsius has an arbitrary zero point (0°C does not mean total absence of heat), so it is an Interval scale. Ratios like 'twice as hot' are not meaningful on this scale.
Week 1 Diagnostic: Is the number of students attending a lecture a discrete or continuous numerical variable? :: Answer: A — Student counts take distinct non-negative integer values (0, 1, 2, ...), making it a discrete numerical variable.

Practice & Self-Check

Attempt these diagnostic questions to verify your operational understanding.

Question 1

Which scale of measurement applies to 'temperature measured in Celsius'?

MCQ

3 Unit Assessment

A. Nominal

B. Ordinal

C. Interval

D. Ratio

Question 2

Is the number of students attending a lecture a discrete or continuous numerical variable?

MCQ

3 Unit Assessment

A. Discrete

B. Continuous

C. Categorical nominal

D. Categorical ordinal