RESEARCH METHODS

Design, Sampling, Data & Analysis

Measurement & Scaling • Questionnaire Design • Sampling Methods

Primary & Secondary Data • Chi-Square • t-Test • ANOVA
Correlation • Simple & Multiple Regression • Report Writing
Research Methods – Complete Study Notes

UNIT 2
Research Methods: Design, Sampling, and Data

2.1 Measurement and Scaling

Measurement is the process of assigning numbers or labels to objects, events, or people
according to specific rules. Scaling is creating a continuum (a line or spectrum) on which
measured objects are located.

Simple Analogy: Measurement is like taking someone’s temperature with a thermometer.

Scaling is like creating the temperature markings on the thermometer itself. You need the scale
first, then you measure.

A) Scales of Measurement
There are four fundamental scales, arranged from lowest to highest level of precision.
Remember the acronym NOIR – Nominal, Ordinal, Interval, Ratio.

📊 Hierarchy of Measurement Scales

Level 4 (Highest): RATIO ← True zero + all math possible
Level 3: INTERVAL ← Equal gaps + no true zero
Level 2: ORDINAL ← Rank/order matters
Level 1 (Lowest): NOMINAL ← Names/labels only

Each higher level includes ALL features of lower levels

1. Nominal Scale (Naming)

What: Labels or categories with no order. Numbers are used only as IDs, not quantities.

💡 Real-Life Example: Jersey Numbers & Gender

Cricket jersey numbers: Virat = 18, Dhoni = 7 – these are just labels. 18 is NOT better than 7.
Similarly: Gender (Male=1, Female=2), Blood group (A, B, AB, O), Pin codes (395001, 110001).
You can only COUNT how many fall in each category (mode).

2. Ordinal Scale (Ordering)

What: Categories WITH meaningful order, but gaps between ranks are NOT equal.

Page 2
Research Methods – Complete Study Notes

💡 Real-Life Example: Customer Satisfaction

Very Unhappy (1) < Unhappy (2) < Neutral (3) < Happy (4) < Very Happy (5). We know Happy >
Unhappy, but is the gap between 1 and 2 the same as between 4 and 5? We don’t know!
Similarly: Education level (High school < Bachelor’s < Master’s < PhD), Army ranks.

3. Interval Scale (Equal Intervals)

What: Order + equal gaps between values, but NO true zero point.

💡 Real-Life Example: Temperature & Calendar Years

20°C to 30°C = same gap as 30°C to 40°C. But 0°C doesn’t mean “no temperature.” You can’t
say 40°C is “twice as hot” as 20°C. Calendar years: the gap from 2000 to 2010 = 2010 to 2020,
but Year 0 ≠ “no time.” IQ scores are also interval.

4. Ratio Scale (Real Zero)

What: Highest level. Has order, equal intervals, AND a true zero (zero = nothing). All math
operations work.

💡 Real-Life Example: Weight, Height, Income, Sales

Weight: 0 kg = no weight. 80 kg is truly 2× of 40 kg. Income: ₹0 = no income, ₹1 lakh is exactly
double of ₹50,000. Age: 0 years = birth, 30 years is 3× of 10 years. All +, –, ×, ÷ operations are
valid.

Feature Nominal Ordinal Interval Ratio

Categorize? ✔ Yes ✔ Yes ✔ Yes ✔ Yes

Rank/Order? ✘ No ✔ Yes ✔ Yes ✔ Yes

Equal gaps? ✘ No ✘ No ✔ Yes ✔ Yes

True zero? ✘ No ✘ No ✘ No ✔ Yes

Allowed stats Mode, Count, % Median, Mean, SD, All (incl. ratios,
Percentiles Correlation GM)

Example Gender, Color Rankings, Grades Temperature °C, Weight kg, Age,
IQ Income

B) Criteria for Good Measurement

A measurement instrument (like a questionnaire or a scale) must meet certain quality standards
before it can be trusted. The three main criteria are:

Page 3
Research Methods – Complete Study Notes

1. Reliability (Consistency)
If you measure the same thing multiple times, do you get the same result? Like a weighing scale
that always shows 70 kg for the same person.

Type of Reliability What It Checks Example

Test-Retest Same test → same people → IQ test in January & March gives
different times → similar results? similar scores

Internal Consistency Do all items in a scale measure the All 10 questions on “job satisfaction”
(Cronbach’s α) same thing? correlate (α > 0.7 = good)

Inter-Rater Different observers give similar Two judges score the same essay
ratings? similarly

Split-Half Split items into halves → both halves Odd-numbered vs even-numbered

give similar scores? questions correlate

2. Validity (Accuracy)
Does the instrument actually measure what it’s supposed to measure? A valid thermometer
measures temperature, not humidity.

Type of Validity What It Checks Example

Content Validity Does it cover ALL aspects of the Marketing exam should cover 4Ps,
concept? STP, branding – not just pricing

Construct Validity Does it measure the theoretical An “anxiety scale” should measure
concept correctly? anxiety, not depression

Criterion Validity Does it correlate with an external SAT scores predict college GPA →
standard? SAT has criterion validity

Face Validity Does it LOOK like it measures what it A “math test” should look like math,
should? (Weakest) not history

3. Sensitivity
Can the instrument detect small differences? A bathroom scale that measures only in 5 kg
increments cannot detect a 1 kg weight loss – it lacks sensitivity. A good scale should be
precise enough for your purpose.

📊 Reliability vs Validity – Dart Board Analogy

Reliable but Valid but Both Reliable
NOT Valid NOT Reliable AND Valid

Page 4
Research Methods – Complete Study Notes

O O O O
O O O O O O O
O O

Same spot each Near center, but Hits center

time, but WRONG scattered every time =
spot randomly PERFECT!

📝 Key Takeaway: A study can be reliable without being valid (a broken clock is consistently
wrong), but a valid study MUST be reliable. Validity is the HIGHER standard.

C) Factors in Selecting a Measurement Scale

When choosing which scale to use in your research, consider these factors:

Factor What to Consider Example

Research Objective What exactly do you want to measure? If just classifying: Nominal. If
Simple categories or precise quantities? measuring exact amounts:
Ratio

Nature of the Variable Is it qualitative (categories) or quantitative Religion = Nominal, Income =

(numbers)? Ratio, Satisfaction = Ordinal

Level of Precision How detailed must the measurement be? For rough grouping: Ordinal.
Needed For exact analysis:
Interval/Ratio

Statistical Techniques What analysis will you do? Higher scales Mean & regression need
allow more powerful tests Interval/Ratio; Chi-square
works with Nominal

Respondent’s Ability Can respondents handle the scale Children: simple Yes/No.
complexity? Adults: 7-point Likert scale

Time & Cost More precise scales need more careful A simple ranking is cheaper
design and time than calibrating a ratio
instrument

2.2 Questionnaire Design

A) Meaning of Questionnaire

Page 5
Research Methods – Complete Study Notes

A questionnaire is a structured set of written questions given to respondents to collect data. It is

the most common data collection tool in survey research. It can be administered in person, by
mail, online, or by phone.

Think of it as: A conversation between researcher and respondent, written on paper. The
quality of your answers depends on the quality of your questions!

Feature Description
Open-ended Questions Respondent answers in their own words. e.g., “What do you think
about our service?”

Closed-ended Questions Fixed choices are given. e.g., “Rate our service: Excellent / Good /
Average / Poor”

Dichotomous Questions Only 2 choices: Yes/No, True/False, Male/Female

Multiple Choice Several options, pick one or more

Likert Scale Strongly Agree → Strongly Disagree (5 or 7 points)

Ranking Questions Put items in order of preference (1st, 2nd, 3rd...)

B) Phases of Questionnaire Design

Designing a good questionnaire is a multi-step process. Rushing through it leads to bad data.

📊 Phases of Questionnaire Design

Phase 1 Phase 2 Phase 3
┌─────────────┐ ┌─────────────┐ ┌─────────────┐
│ Define │→→│ Determine │→→│ Draft │
│ Information │ │ Question │ │ Questions │
│ Needed │ │ Type & Form │ │ │
└─────────────┘ └─────────────┘ └─────────────┘
▼ ▼ ▼
Phase 4 Phase 5 Phase 6
┌─────────────┐ ┌─────────────┐ ┌─────────────┐
│ Decide │→→│ Arrange │→→│ Pilot Test │
│ Sequence & │ │ Layout & │ │ & Revise │
│ Flow │ │ Formatting │ │ │
└─────────────┘ └─────────────┘ └─────────────┘

Page 6
Research Methods – Complete Study Notes

Phase Activity Details

Phase 1: Define Info What do you want to List research objectives → translate each into
Needed learn? measurable questions

Phase 2: Question Open, closed, Likert, Match question type to the type of data needed
Type ranking?

Phase 3: Draft Write initial questions Use simple language, one idea per question,
Questions avoid leading/double-barreled

Phase 4: Sequence & Arrange in logical order Easy → difficult. General → specific.
Flow Demographics last (or first if easy warm-up)

Phase 5: Layout & Make it visually appealing Clear headings, consistent fonts, enough space,
Format professional appearance

Phase 6: Pilot Test Test on 15-30 people first Find confusing questions, typos, measure
completion time. Revise based on feedback

Golden Rules of Question Writing

Rule ✘ Bad Example ✔ Good Example
Keep it simple “What is the frequency of your “How often do you exercise?”
physical exertion regimen?”

One idea per “Do you like the food AND service Ask food & service as 2 separate
question (no double- here?” questions
barreled)

Avoid leading “Don’t you agree our product is “How would you rate our product?”
questions amazing?”

Avoid “Do you support wasteful government “What is your opinion on current
loaded/emotional spending?” government spending?”
words

Provide balanced “Good / Very Good / Excellent” (all “Very Poor / Poor / Neutral / Good /
options positive!) Very Good”

Give clear time frame “Do you exercise regularly?” (vague) “How many times did you exercise
LAST WEEK?”

💡 Real-Life Example: Good vs Bad Questionnaire Design

Bad questionnaire for a restaurant: Q1: Don’t you think our pizza is delicious? (Leading!) Q2:
How is the food and ambience and parking? (Triple-barreled!) Good questionnaire: Q1: On a
scale of 1-5, how would you rate the taste of our pizza? Q2: On a scale of 1-5, how would you
rate the restaurant ambience? Q3: On a scale of 1-5, how satisfied are you with parking
availability?

Page 7
Research Methods – Complete Study Notes

2.3 Sampling and Sampling Distribution

A) Essentials of Sampling
Sampling is the process of selecting a subset (sample) from a larger group (population) to
study. Since studying the entire population is usually impossible (too expensive, too slow), we
study a representative sample instead.

📊 Key Sampling Terms

┌───────────────────────────────────────┐
│ POPULATION: All students in India │
│ (Too large to study everyone) │
│ │
│ ┌─────────────────────┐ │
│ │ SAMPLE: 1,000 │ │
│ │ selected students │ │
│ └─────────────────────┘ │
│ │
│ Sampling Frame = List from which we │
│ actually pick (e.g., university DB) │
└───────────────────────────────────────┘

B) Sampling Design Process

📊 Steps in Sampling Design
Step 1: Define the Target Population
(Who exactly do you want to study?)
▼
Step 2: Determine the Sampling Frame
(What list will you pick from?)
▼
Step 3: Select Sampling Method
(Probability or Non-probability?)
▼
Step 4: Determine Sample Size
(How many people to include?)
▼
Step 5: Execute the Sampling
(Actually select the participants)

Page 8
Research Methods – Complete Study Notes

C) Random (Probability) Sampling Methods

In probability sampling, every member of the population has a KNOWN, non-zero chance of
being selected. Results can be generalized to the population.

Method How It Works Example Best For

Simple Random Every person has equal Assign numbers to 5,000 Homogeneous
chance (lottery) students, use random populations; small-
generator to pick 200 medium size

Stratified Divide into groups Divide employees by When subgroups are

Random (strata), random pick department, then important (age, gender)
from each randomly pick from each

Cluster Divide into clusters Randomly pick 50 Large geographic areas;

(areas), randomly pick schools, survey ALL reduces travel cost
some clusters, study all students in them
in them

Systematic Pick every k-th item from From 10,000 names, pick Ordered lists; quick to
a list every 50th name (k=200) execute

💡 Real-Life Example: Stratified Sampling in Action

An IT company wants to survey employee satisfaction. They have 60% engineers, 25%
managers, 15% support staff. For a sample of 200: • Engineers: 200 × 0.60 = 120 randomly
selected • Managers: 200 × 0.25 = 50 randomly selected • Support staff: 200 × 0.15 = 30
randomly selected This ensures every department’s voice is heard proportionally!

D) Non-Random (Non-Probability) Sampling

Not everyone has an equal chance of selection. Quicker and cheaper, but results may not be
generalizable.

Method How It Works Example Limitation

Convenience Survey whoever is easily A student surveys friends Highly biased; not
available on campus representative

Judgmental Researcher picks “best” Interviewing only senior Depends on researcher’s

(Purposive) subjects using expertise doctors for a healthcare judgment
study

Quota Set quotas per category, Need 50 males + 50 No randomness within

fill by convenience females; approach quotas
people until filled

Page 9
Research Methods – Complete Study Notes

Snowball Existing participants refer Studying illegal drug use Only reaches connected
new ones – one user refers others networks

E) Sampling and Non-Sampling Errors

Feature Sampling Error Non-Sampling Error

What is it? Difference between sample result Mistakes from poor design, data
and true population value collection, or processing

Why it happens Because we study only a PART of Human mistakes, biased questions,
the population wrong data entry, non-response

Can it be measured? Yes – using standard error, No – very difficult to quantify

confidence intervals

How to reduce? Increase sample size; use Better training, pilot testing, careful
probability sampling design, double-checking data

Happens in census? No (you study everyone) Yes! Even a census can have non-
sampling errors

Formula-related? SE = σ/√n (decreases as n No formula; prevention is the cure

increases)

💡 Real-Life Example: Sampling vs Non-Sampling Error

Sampling error: You survey 100 people about average income and get ₹45,000. The true
population average is ₹48,000. The ₹3,000 gap is sampling error – it exists simply because you
didn’t ask everyone. Non-sampling error: In the same survey, some people lie about their
income (response bias), 20% don’t respond (non-response bias), and a data entry clerk types
₹45,000 as ₹450,000 (processing error). These are all non-sampling errors.

2.4 Sources and Collection of Data

A) Primary vs Secondary Data

📊 Types of Data
┌───────────────────────────────┐
│ DATA SOURCES │
└─────────────┬─────────────────┘
│

Page 10
Research Methods – Complete Study Notes

┌───────┴───────┐ ┌──────────────┐
│ PRIMARY DATA │ │ SECONDARY DATA │
│ (Original, │ │ (Already │
│ first-hand) │ │ collected) │
└───────────────┘ └──────────────┘
Qualitative: Internal:
- Observation - Company records
- Interview - Sales reports
Quantitative: External:
- Survey - Govt databases
- Experiment - Published research

Primary Data: Data collected by the researcher themselves, directly from the source, for their
specific research problem. It is original and fresh.
Secondary Data: Data that was collected earlier by someone else for a different purpose but is
now being reused by the researcher.

B) Benefits and Limitations of Secondary Data

Benefits (✔ Advantages) Limitations (✘ Disadvantages)
Saves time – already available; no need to collect May be outdated – data could be years old

Saves money – much cheaper than primary May not fit your needs – collected for different
research purpose

Large datasets available (Census, World Bank) Accuracy unknown – you don’t know the
collection quality

Good for understanding trends over time Definition differences – “income” might be defined
differently

Can compare across regions/countries easily No control over methodology used in collection

Available for sensitive/hard-to-research topics May be biased by original collector’s agenda

C) Roadmap to Use Secondary Data

Before blindly using secondary data, follow this systematic roadmap:

📊 Steps to Evaluate Secondary Data

Step 1: What was the PURPOSE of the original study?

Page 11
Research Methods – Complete Study Notes

(Does it align with your research question?)

▼
Step 2: WHO collected the data?
(Credible source? Government, university, reputed org?)
▼
Step 3: HOW was it collected?
(What methodology, sample size, sampling method?)
▼
Step 4: WHEN was it collected?
(Is it recent enough to be relevant?)
▼
Step 5: Is the data CONSISTENT?
(Cross-verify with other sources if possible)
▼
Step 6: Does the data FIT your definitions?
(Are variables defined the same way you need?)

D) Secondary Data Sources

Source Type Examples Type of Data
Government Census, RBI Bulletin, Economic Survey, Population, economic, financial,
Publications NSSO reports social

International World Bank, IMF, WHO, UNESCO Cross-country comparisons,

Organizations datasets development indicators

Industry Reports McKinsey, Deloitte, KPMG, IBEF sector Market size, trends,
reports competition

Academic Journals JSTOR, Google Scholar, SSRN, PubMed Research findings, literature
review

Company Records Annual reports, financial statements, Sales, revenue, employee data
internal databases

Media & Digital Newspapers, websites, social media Opinions, trends, sentiment
analytics, Google Trends

E) Primary Data Collection Methods

QUALITATIVE Methods (Words, Themes, Insights)

Page 12
Research Methods – Complete Study Notes

1. Observation: The researcher watches and records behavior without asking questions. Can
be participant (researcher joins the group) or non-participant (researcher watches from outside).

💡 Real-Life Example: Observation in Retail

A retail company hires researchers to stand in their store and observe: Where do customers go
first? How long do they browse? What do they touch but not buy? This observational data helps
redesign store layout. The customers don’t know they’re being studied (non-participant
observation).

Feature Participant Observation Non-Participant Observation

Researcher’s role Actively participates in the group Watches from outside; no
participation

Depth of insight Very deep – experiences from inside Surface level but more objective

Bias risk High – may “go native” and lose Lower – maintains distance
objectivity

Example Researcher works at a factory to Researcher watches children playing

study worker conditions in a park from a bench

2. Interview: A face-to-face or virtual conversation where the researcher asks questions and
records answers. Can be structured (fixed questions), semi-structured (guided but flexible), or
unstructured (free-flowing conversation).

Type Structure When to Use Example

Structured Pre-set questions in Large sample, easy Phone survey with 20
fixed order comparison standard questions

Semi-Structured Key questions + room Moderate depth + some Interviewing managers

for follow-up standardization about leadership styles

Unstructured / In- Open conversation, no Exploring unknown Talking to tribal elders about
depth fixed questions topics deeply cultural practices

QUANTITATIVE Methods (Numbers, Measurement)

1. Survey (Most Common): A standardized questionnaire administered to a large number of

respondents. Can be done online, face-to-face, by phone, or by mail.

Survey Mode Pros Cons Best For

Online (Google Fast, cheap, wide Low response rate, Young/tech-savvy populations
Forms, reach, easy data internet needed
SurveyMonkey) entry

Page 13
Research Methods – Complete Study Notes

Face-to-Face Highest response Expensive, time- Illiterate populations, sensitive

rate, can clarify consuming, topics
doubts interviewer bias

Telephone Moderate cost, quick, No visual aids, Quick opinion polls, follow-ups
good response declining landline use

Mail/Postal No interviewer bias, Very low response Rarely used today; rural areas
respondent comfort rate (10-20%), slow

2. Experimentation: The researcher manipulates one variable (independent) to measure its

effect on another (dependent), while controlling all other variables. It is the ONLY method that
can establish cause-and-effect.

💡 Real-Life Example: Experimentation – A/B Testing

Netflix wants to know if changing the thumbnail image for a show increases clicks. They show
Image A to 50,000 users and Image B to another 50,000 (randomly assigned). After 2 weeks:
Image A got 12% clicks, Image B got 18% clicks. Conclusion: Image B CAUSED more clicks.
This is a controlled experiment (A/B test).

Feature Survey Experiment

Purpose Describe, measure, explore Establish cause-and-effect

Control over variables Low – no manipulation High – researcher controls variables

Setting Natural, real-world Controlled (lab or field)

Cause-effect? Cannot establish causation CAN establish causation

Example Customer satisfaction survey A/B testing on website

Cost & Time Moderate High (requires careful control)

📝 Key Takeaway: Use QUALITATIVE methods (observation, interviews) when exploring a new
topic. Use QUANTITATIVE methods (survey, experiment) when testing a specific hypothesis
with numbers.

Page 14
Research Methods – Complete Study Notes

UNIT 3
Data Analysis and Presentation

3.1 Hypothesis Testing for Categorical Data

A) Chi-Square Test (χ²)

The Chi-Square test checks whether there is a significant association between two categorical
(nominal) variables. It compares OBSERVED frequencies with EXPECTED frequencies.

📐 Chi-Square Formula

χ² = Σ (O - E)² / E

O = Observed frequency (what actually happened)

E = Expected frequency (what we’d expect if no association)
E = (Row Total × Column Total) / Grand Total
df = (rows - 1) × (columns - 1)

Decision: If χ² calculated > χ² critical → REJECT H₀

💡 Real-Life Example: Chi-Square: Gender vs Online Shopping Preference

Survey of 200 people: Prefer Online Prefer Offline Total Male 65
35 100 Female 45 55 100 Total 110 90
200 Expected (if no association): E(Male,Online) = (100×110)/200 = 55 E(Male,Offline) =
(100×90)/200 = 45 E(Female,Online) = (100×110)/200 = 55 E(Female,Offline) = (100×90)/200
= 45 χ² = (65-55)²/55 + (35-45)²/45 + (45-55)²/55 + (55-45)²/45 = 1.82 + 2.22 + 1.82 + 2.22 =
8.08 df = 1, Critical value = 3.841 at α=0.05 Since 8.08 > 3.841 → REJECT H₀ Conclusion:
Gender and shopping preference ARE significantly associated. Males tend to prefer online
shopping more.

B) t-Test
The t-test compares means between groups when sample sizes are small and/or population
standard deviation is unknown.

Type When to Use Example

Page 15
Research Methods – Complete Study Notes

One-sample t-test Compare sample mean to a known Is our class average = national
value average of 75?

Independent samples t- Compare means of 2 separate Do males and females differ in

test groups spending?

Paired t-test Compare same group before & Did a training program improve
after scores?

📐 Independent Samples t-Test

̄
X₁ - ̄
X₂
t = ────────────────────
√(S²₁/n₁ + S²₂/n₂)

̄X₁, ̄
X₂ = sample means
S²₁, S²₂ = sample variances
n₁, n₂ = sample sizes

💡 Real-Life Example: t-Test: Male vs Female Spending

A researcher compares monthly shopping spend: Males (n=30): Mean = ₹5,200, SD = ₹1,100
Females (n=30): Mean = ₹6,800, SD = ₹1,300 t = (5200-6800) / √(1100²/30 + 1300²/30) = -
1600 / √(40333 + 56333) = -1600 / 310.9 = -5.15 |t| = 5.15 > 2.001 (critical at df=58, α=0.05)
REJECT H₀: Females spend significantly more than males on shopping.

C) ANOVA (Analysis of Variance)

ANOVA compares means across THREE or more groups simultaneously. It uses the F-statistic
to test whether at least one group mean is significantly different.

📐 One-Way ANOVA

MSB Mean Square Between Groups

F = ──── = ───────────────────────────
MSW Mean Square Within Groups

Large F → Groups differ significantly

Small F → No significant difference

Page 16
Research Methods – Complete Study Notes

💡 Real-Life Example: ANOVA: Comparing 3 Training Methods

A company tests 3 training programs on employee productivity: Method A: Mean score = 72
Method B: Mean score = 78 Method C: Mean score = 85 ANOVA Result: F = 6.42, p-value =
0.003 Since p < 0.05 → REJECT H₀ Conclusion: At least one training method produces
significantly different results. Next step: Use Tukey’s post-hoc test to find WHICH methods
differ (result: C is significantly better than A, but B is not significantly different from either).

📊 Choosing the Right Test

What type of data do you have?
│
┌─────┴──────┐
CATEGORICAL NUMERICAL
│ │
χ² TEST How many groups?
│
┌─────┴──────┐
2 groups 3+ groups
│ │
t-TEST ANOVA

3.2 Correlation and Simple Linear Regression

A) Correlation Analysis
Correlation measures the strength and direction of the linear relationship between two numerical
variables. The Pearson correlation coefficient (r) ranges from -1 to +1.

📊 Interpreting Correlation (r)

r = +1.0 Perfect positive (both rise together)
r = +0.7 Strong positive
r = +0.3 Weak positive
r = 0.0 No linear relationship
r = -0.3 Weak negative
r = -0.7 Strong negative
r = -1.0 Perfect negative (one up, other down)

GOLDEN RULE: Correlation ≠ Causation!

Page 17
Research Methods – Complete Study Notes

📐 Pearson’s r

nΣXY - (ΣX)(ΣY)
r = ────────────────────────────────
√[nΣX² - (ΣX)²][nΣY² - (ΣY)²]

B) Simple Linear Regression

Regression goes beyond correlation – it gives you a PREDICTION EQUATION. If you know X,
you can predict Y.

📐 Simple Linear Regression

Y = a + bX

b (slope) = nΣXY - (ΣX)(ΣY)

─────────────────
nΣX² - (ΣX)²

a (intercept) = ̄
Y - b̄X

r² = Coefficient of Determination
= % of variation in Y explained by X

💡 Real-Life Example: Regression: Ad Spend Predicting Sales

Data: X = Ad spend (₹ lakhs), Y = Sales (₹ lakhs) Regression result: Y = 15.2 + 4.1X, r² = 0.89
Interpretation: • Every ₹1 lakh extra spent on ads increases sales by ₹4.1 lakhs • With zero
advertising, base sales = ₹15.2 lakhs • 89% of sales variation is explained by ad spending
Prediction: If ad spend = ₹10 lakhs: Sales = 15.2 + 4.1(10) = 15.2 + 41 = ₹56.2 lakhs

3.3 Multiple Regression Analysis

Simple regression uses ONE independent variable (X) to predict Y. But in real life, outcomes
depend on MANY factors. Multiple regression uses TWO or more independent variables
simultaneously.

Page 18
Research Methods – Complete Study Notes

📐 Multiple Regression Equation

Y = a + b₁X₁ + b₂X₂ + b₃X₃ + ... + bkXk

Where:
Y = Dependent variable (what you predict)
X₁, X₂, X₃... = Independent variables (predictors)
b₁, b₂, b₃... = Regression coefficients (impact of each X)
a = Intercept (Y when all X’s = 0)

R² = Multiple Coefficient of Determination

= % of Y’s variation explained by ALL X’s together
Adjusted R² = Penalizes for adding weak variables

💡 Real-Life Example: Multiple Regression: Predicting House Prices

A real estate analyst predicts house prices (Y) using 3 factors: X₁ = Area (sq ft), X₂ = Number
of bedrooms, X₃ = Distance from city center (km) Result: Price = 5,00,000 + 3,000(Area) +
8,00,000(Bedrooms) - 50,000(Distance) R² = 0.92 Interpretation: • Each extra sq ft adds
₹3,000 to price • Each extra bedroom adds ₹8 lakhs • Each km farther from city reduces price
by ₹50,000 • 92% of price variation is explained by these 3 factors Prediction: A 1,200 sq ft, 3
BHK flat, 5 km from center: Price = 5,00,000 + 3,000(1200) + 8,00,000(3) - 50,000(5) Price =
5,00,000 + 36,00,000 + 24,00,000 - 2,50,000 = ₹63,50,000

Simple vs Multiple Regression

Feature Simple Regression Multiple Regression
Predictors 1 independent variable (X) 2 or more (X₁, X₂, ...)

Equation Y = a + bX Y = a + b₁X₁ + b₂X₂ + ...

When to use One factor clearly dominates Multiple factors influence the
outcome

Example Ad spend → Sales Ad spend + price + season → Sales

r² interpretation % explained by one variable R² = % explained by ALL variables

together

Risk Omitted variable bias (missing Multicollinearity (predictors correlated

important factors) with each other)

Key Concept: Multicollinearity

Page 19
Research Methods – Complete Study Notes

When two independent variables are highly correlated with EACH OTHER (e.g., height and
weight both predicting blood pressure), it creates problems. The model can’t separate their
individual effects. Solution: remove one redundant variable or use techniques like VIF (Variance
Inflation Factor) to detect it.

📝 Key Takeaway: Always check R² AND Adjusted R². R² always increases when you add more
variables (even useless ones). Adjusted R² penalizes for unnecessary variables, so it’s a better
measure of model quality.

3.4 Tools for Primary and Secondary Data Analysis

Tool / Software Type Best For Example Use

SPSS Statistical Surveys, social science, Running t-tests, ANOVA,
business research regression on survey data

Excel / Google Spreadsheet Basic analysis, charting, Descriptive stats, simple

Sheets pivot tables charts, data cleaning

R / R Studio Programming Advanced statistics, Multiple regression, time

custom analysis series, complex models

Python (Pandas, Programming Data science, machine Large dataset analysis,

SciPy) learning automation, visualization

Stata Statistical Econometrics, panel data Regression analysis for

economics research

NVivo / [Link] Qualitative Interview transcripts, Coding and finding themes in

thematic analysis interview data

Tableau / Power BI Visualization Dashboards, presentations Creating interactive charts for

reports

Google Analytics Digital Website/app behavior Tracking user behavior,

analysis conversion rates

Choosing the Right Tool

📊 Decision Guide for Analysis Tools
What type of data?
│
┌───┴─────────┬───────────────┬─────────────┐
Small & Medium & Large / Qualitative

Page 20
Research Methods – Complete Study Notes

Simple Structured Complex (text/audio)

│ │ │ │
EXCEL SPSS / Stata Python / R NVivo
[Link]

3.5 Layout of a Research Report

A research report is the final deliverable of your study. It communicates your findings to the
reader in a structured, professional format. Think of it as telling the story of your research
journey.

📊 Standard Research Report Structure

┌───────────────────────────────────────┐
│ PRELIMINARY PAGES (Front Matter) │
│ Title Page → Abstract → TOC → │
│ List of Tables/Figures │
├───────────────────────────────────────┤
│ MAIN BODY │
│ Ch 1: Introduction & Problem │
│ Ch 2: Literature Review │
│ Ch 3: Research Methodology │
│ Ch 4: Data Analysis & Results │
│ Ch 5: Discussion & Interpretation │
│ Ch 6: Conclusions & Recommendations │
├───────────────────────────────────────┤
│ END MATTER (Back Matter) │
│ References → Appendices → │
│ Questionnaire copy │
└───────────────────────────────────────┘

Detailed Layout
Section What to Include Tips
Title Page Title, author name, institution, date, Make title specific: “Impact of
supervisor’s name Social Media on Purchase
Decisions of Gen Z in Surat”

Abstract Summary of entire study in 150-300 words: Write this LAST, even though it
problem, method, key findings, conclusion appears first!

Table of Chapter headings with page numbers Auto-generate using MS Word

Page 21
Research Methods – Complete Study Notes

Contents

Ch 1: Background, problem statement, research Hook the reader – why does this
Introduction objectives, hypotheses, significance, scope study matter?
& limitations

Ch 2: Literature Summary of existing research, theoretical Organize by themes, not just

Review framework, research gap you are filling author-by-author

Ch 3: Research design, population, sample, Be specific: “384 respondents

Methodology sampling method, data collection tool, data using stratified random sampling”
analysis techniques

Ch 4: Data Tables, charts, statistical test results (t-test, Present data FIRST, then interpret;
Analysis ANOVA, regression, etc.) use visuals

Ch 5: Discussion Compare findings with literature review; Link back to research objectives
explain why results occurred and hypotheses

Ch 6: Conclusion Summary of findings, practical Answer: “So what? What should

recommendations, limitations, future be done now?”
research suggestions

References All cited works in APA/Harvard format Use Zotero or Mendeley to

manage references

Appendices Questionnaire copy, raw data tables, Number them: Appendix A,

consent forms, approval letters Appendix B, etc.

Tips for Writing an Excellent Report

Do’s (✔) Don’ts (✘)
Write in third person (“The researcher found...”) Don’t use first person (“I found...”) unless
instructed

Use past tense for methodology and results Don’t switch between past and present tense
randomly

Present data in tables AND explain in text Don’t put a table without interpreting it

Be specific: “72% of respondents agreed” Don’t be vague: “Many people agreed”

Cite every claim that’s not your own finding Don’t plagiarize – always give credit

Proofread for grammar and formatting Don’t submit without spell-checking

Include limitations honestly Don’t hide weaknesses – acknowledge them

Keep it concise and focused Don’t pad with irrelevant content to increase
pages

💡 Real-Life Example: Sample Abstract

Page 22
Research Methods – Complete Study Notes

This study investigates the impact of social media marketing on purchase decisions among Gen
Z consumers in Surat. A structured questionnaire was administered to 384 respondents selected
through stratified random sampling. Data was analyzed using descriptive statistics, Chi-square
test, and multiple regression. Results indicate that Instagram ads (β=0.42, p<0.01) and
influencer reviews (β=0.31, p<0.01) significantly influence purchase intent, while Facebook ads
showed no significant effect. The model explained 67% of variance (R²=0.67). The study
recommends that companies targeting Gen Z should prioritize Instagram and influencer
partnerships over traditional Facebook advertising.

📝 Key Takeaway: The abstract is the MOST READ section of any research paper. Keep it
concise, informative, and include: problem, method, key findings, and conclusion – all in about
200 words.

Page 23
Research Methods – Complete Study Notes

Quick Revision Summary

Topic Key Point Remember This

Nominal Scale Labels/categories, no order Gender, Blood type, Pin code

Ordinal Scale Order exists, unequal gaps Ranks, Satisfaction ratings

Interval Scale Equal gaps, no true zero Temperature °C, IQ, Calendar year

Ratio Scale True zero, all math works Weight, Income, Age, Height

Reliability Consistency of measurement Cronbach’s α > 0.7 = good

Validity Accuracy of measurement Content, Construct, Criterion, Face

Questionnaire Design 6 phases: Define → Type → Draft → Always pilot test before finalizing!
Sequence → Layout → Pilot

Probability Sampling Random: Simple, Stratified, Cluster, Results are generalizable

Systematic

Non-Probability Non-random: Convenience, Purposive, Fast but may not be generalizable

Sampling Quota, Snowball

Sampling Error Gap between sample & population (use SE = σ/√n

larger n)

Non-Sampling Error Human mistakes, bias, non-response Better design & training reduce it

Primary Data Collected by you: Survey, Experiment, Expensive but specific

Interview

Secondary Data Already exists: Census, Reports, Cheap but may not fit perfectly
Journals

Chi-Square Association between 2 categorical χ² = Σ(O-E)²/E

variables

t-Test Compare 2 group means Paired (before-after) or

Independent (2 groups)

ANOVA Compare 3+ group means F = MSB/MSW; use post-hoc if

significant

Correlation (r) Strength & direction of relationship -1 ≤ r ≤ +1; Correlation ≠

Causation

Simple Regression Y = a + bX (one predictor) r² = % of Y explained by X

Multiple Regression Y = a + b₁X₁ + b₂X₂ +... Check Adjusted R² and

multicollinearity

Page 24
Research Methods – Complete Study Notes

Research Report Prelim → Body → End matter Abstract is most-read section; write
it last!

─── End of Notes ───

Page 25