Business Research Methods — Module 3: Data Collection & Processing
Lecture Notes • Complete Study Material
13. Sources of Data Collection: Primary vs. Secondary Sources
Data collection is the process of gathering information relevant to a research problem. The quality and credibility of research findings largely depend on the validity and reliability of the data collected.
Primary Data
Original data collected by the researcher for the first time specifically for the purpose of the study, directly from respondents or live observation.
- Characteristics: Original, highly specific, more accurate, resource-intensive.
- Advantages: Tailored to exact objectives, complete control over collection methodology.
- Disadvantages: Expensive and time-consuming.
Secondary Data
Data already collected and published by other agencies or researchers for another purpose, utilized by the researcher as background or comparative input.
- Sources: Internal (sales reports, financial statements) and External (Census, journals, govt databases).
- Advantages: Economical, quickly accessible, ideal for preliminary exploratory reviews.
- Disadvantages: May not fit specific study objectives; risk of outdated metrics.
| Primary Data | Secondary Data |
|---|---|
| Collected first-hand directly from respondents | Already collected and curated by others |
| Highly tailored to the exact research problem | May be less specific to current problem goals |
| Expensive and time-consuming | Economical and quickly accessible |
| High accuracy and freshness control | Accuracy is entirely dependent on original source |
Primary Data Collection Methods
1. Observation Method
Collecting empirical data by watching behaviors, events, or situations directly without asking questions:
- Participant vs Non-Participant: Researcher joins group vs observes from outside.
- Structured vs Unstructured: Predefined observational checklist vs open flexible logging.
- Pros & Cons: Direct real-world behavior capture; susceptible to observer bias.
2. Interview Method
Collecting verbal responses through direct interactive questioning between researcher and respondent:
- Structured vs Unstructured: Fixed question script vs open exploratory dialogue.
- Personal vs Telephone: Face-to-face interaction vs remote telephonic interviews.
- Pros & Cons: In-depth clarification; higher cost and potential interviewer bias.
3. Questionnaire Method
A standardized written instrument consisting of structured questions completed independently by respondents.
- Characteristics: Low administration cost, suitable for large geographically dispersed samples.
- Limitations: Lower response rates and potential misinterpretation of questions.
4. Schedule Method
A formal list of questions administered and filled in directly by a trained enumerator in person.
- Characteristics: High completion rate, ensures responses even from illiterate populations (e.g., National Census).
- Limitations: Higher operational expense; requires trained field investigators.
14. Design and Development of Questionnaires
A well-designed questionnaire ensures respondent cooperation and maximizes data accuracy while eliminating ambiguity:
Steps in Questionnaire Development
Types of Questions
- Open-Ended: Respondents answer in their own words (e.g., “What improvements do you suggest for our mobile app?”).
- Closed-Ended: Respondents choose from predetermined choices (e.g., Multiple choice, Dichotomous Yes/No, Rating scales).
Characteristics of a Good Questionnaire
- Uses simple, clear, and unambiguous language.
- Follows a logical sequence (general to specific “funnel”).
- Avoids leading, biased, or double-barreled questions.
- Maintains an appropriate, concise length to prevent fatigue.
15 & 16. Measurement Scales & Scaling Techniques
A measurement scale is a mathematical framework used to assign numbers or symbols to characteristics of variables according to rules.
The Four Levels of Measurement Scales
1. Nominal Scale
Classifies data into mutually exclusive categories without order or ranking (e.g., Gender: Male=1, Female=2).
2. Ordinal Scale
Classifies data and provides ranking, but intervals between ranks are unknown (e.g., Satisfaction: High, Medium, Low).
3. Interval Scale
Ranks objects with equal measurable distances between intervals, but lacks a true absolute zero (e.g., Temperature in °C).
4. Ratio Scale
Possesses equal intervals and a true absolute zero point, allowing ratio comparisons (e.g., Sales revenue, Age, Weight).
| Scale | Classification | Ranking | Equal Intervals | True Zero |
|---|---|---|---|---|
| Nominal | Yes | No | No | No |
| Ordinal | Yes | Yes | No | No |
| Interval | Yes | Yes | Yes | No |
| Ratio | Yes | Yes | Yes | Yes |
Scaling Techniques: Comparative vs Non-Comparative
Comparative Scaling
Respondents directly compare two or more stimulus objects against one another:
- Paired Comparison: Comparing two objects at a time.
- Rank Order: Ranking alternatives from most to least preferred.
- Constant Sum: Allocating 100 points across attributes to reflect relative importance.
Non-Comparative Scaling
Respondents evaluate each stimulus object independently on its own merits:
- Likert Scale: 5 or 7-point scale measuring degree of agreement (Strongly Agree to Strongly Disagree).
- Semantic Differential: 7-point rating between bipolar opposite adjectives (e.g., Reliable vs Unreliable).
- Stapel Scale: 10-point unipolar vertical numerical scale ranging from +5 to -5.
17 & 18. Pre-Testing, Pilot Study & Data Processing
Pre-Testing vs. Pilot Study
Pre-Testing
Administering the draft questionnaire to a small group of respondents (15–20) specifically to identify wording flaws, misinterpretations, and question ambiguity before final printing.
Pilot Study
A full-scale rehearsal or trial run of the entire research workflow (sampling, data collection, fieldwork logistics, statistical testing) on a small scale (30–50 respondents) before launching the main investigation.
Processing of Data: 4 Core Stages
Data processing converts raw field responses into clean, structured, and interpretable empirical intelligence:
1. Editing
Inspecting completed survey forms for errors, omissions, legibility, and consistency.
2. Classification
Grouping raw responses into homogeneous classes based on shared characteristics.
3. Coding
Assigning numerical codes or symbols to categories to facilitate computer analysis (e.g., Male=1, Female=2).
4. Tabulation
Arranging coded data into systematic rows and columns in summary tables for statistical testing.
19, 20 & 21. Hypothesis Testing, Errors & Parametric vs. Non-Parametric Tests
Errors in Hypothesis Testing
When evaluating sample data, researchers face the statistical risk of committing two distinct errors:
Type-I Error (α - Alpha)
Rejecting a true null hypothesis (False Positive). For example, concluding that advertising influences sales when in reality it has zero effect. The probability of committing this error is the Level of Significance (typically set at 5% or 0.05).
Type-II Error (β - Beta)
Failing to reject (accepting) a false null hypothesis (False Negative). For example, concluding that advertising does not affect sales when in reality it significantly drives revenue.
Statistical Tests Classification
Parametric Tests
Require assumptions about the underlying population distribution (normal distribution) and metric data measured on interval or ratio scales:
- Z-Test: Comparing sample means when population variance is known and sample size ≥ 30.
- t-Test: Comparing sample means when sample size < 30 and variance is unknown (Student's t-test, Paired t-test).
- ANOVA (F-Test): Comparing variance across three or more group means simultaneously.
Non-Parametric Tests
Distribution-free tests that do not assume a normal distribution; suitable for nominal and ordinal categorical data:
- Chi-Square (χ²) Test: Testing goodness of fit and independence of attributes in contingency tables.
- Mann-Whitney U Test: Non-parametric alternative to independent two-sample t-test.
- Wilcoxon Signed-Rank Test: Non-parametric alternative to paired sample t-test.
22. Use of Computers in Data Processing & Statistical Analysis
Computers play an indispensable role in modern business research by automating storage, complex mathematical computations, statistical testing, and visualization:
- MS Excel: Widely deployed for preliminary data entry, basic descriptive statistics, and charting.
- SPSS (Statistical Package for the Social Sciences): Industry standard for academic and corporate cross-tabulation, ANOVA, regression modeling, and factor analysis.
- R & Python: Advanced programming environments for data science, machine learning models, and big data econometrics.
- Google Forms / Qualtrics: Digital cloud tools for automated multi-channel questionnaire distribution and instant dataset generation.
Download Module 3 Notes (PDF)
Calicut University • FYUGP 2024 Syllabus
Finished this module?
Continue reading the next module or return to the subject overview.