Chapter 1 Introduction to Research Methodology

1.1 Meaning, Objectives, and Types of Research

1.1.1 Definition and Nature of Scientific Research

Scientific research is defined as a systematic, controlled, empirical, and critical investigation of hypothetical propositions regarding presumed relationships among natural and social phenomena (Kothari and Garg (2019)).

Scientific research is distinguished from casual observation by four fundamental characteristics:

  1. Empiricism: Conclusions are strictly grounded in observable, verifiable evidence.
  2. Objectivity: The inquiry is designed to eliminate investigator bias, subjective preferences, and unvalidated assumptions.
  3. Reproducibility: Methodology is documented with sufficient precision to permit exact independent replication.
  4. Systematic Structure: Procedures follow a logical, pre-specified sequence.

1.1.2 Core Objectives of Research

According to Kothari and Garg (2019), the primary purpose of research is to discover answers to questions through the systematic application of scientific procedures. Research objectives are broadly categorized into four types:

  1. Exploratory / Formulative Research: Conducted to gain familiarity with a phenomenon or achieve new insights to formulate a precise research problem or hypothesis (e.g. qualitative focus groups exploring patient perceptions of digital health platforms).
  2. Descriptive Research: Aimed at accurately portraying the characteristics of a particular individual, situation, or population group without manipulating variables (e.g. mapping the demographic distribution and prevalence of hypertension across a municipality).
  3. Diagnostic Research: Focused on determining the frequency with which a phenomenon occurs or evaluating the strength of association between two or more variables (e.g. assessing clinical risk factors associated with 30-day hospital readmission).
  4. Hypothesis-Testing / Analytical Research: Designed to test a formal statistical hypothesis regarding a causal relationship between specific exposure variables and outcomes (e.g. a double-blind randomized clinical trial evaluating whether a novel therapeutic agent reduces disease incidence).

1.1.3 Classification of Research Types

Research in data science, biostatistics, and public health is classified along multiple philosophical and structural dimensions (Creswell and Creswell (2018)):

Dimension Research Type A Research Type B Conceptual Contrast & Empirical Example
Purpose Descriptive Research Analytical Research Descriptive research establishes what exists (e.g. estimating that 25% of patients utilize telehealth). Analytical research evaluates why or how an outcome occurs (e.g. testing whether telehealth adoption reduces emergency department utilization).
Application Basic (Pure) Research Applied Research Basic research seeks to expand foundational theoretical knowledge (e.g. proving convergence properties of an MCMC estimator). Applied research addresses an immediate practical problem (e.g. optimizing intensive care bed allocation algorithms).
Data Paradigm Quantitative Research Qualitative Research Quantitative research relies on numerical measurements and inferential models (e.g. Cox regression of survival times). Qualitative research examines narrative themes and contextual experiences (e.g. semi-structured interviews on clinician burnout).
Environment Empirical (Field) Research Conceptual Research Empirical research gathers observational or experimental field data. Conceptual research formulates abstract frameworks, mathematical definitions, or philosophical models.
Temporal Framework Cross-Sectional Longitudinal Cross-Sectional research captures a single temporal snapshot. Longitudinal research tracks cohort subjects repeatedly across a designated follow-up period.

1.2 Research Methods vs. Research Methodology

A fundamental distinction in research methodology is the difference between Research Methods and Research Methodology (Kothari and Garg (2019)):

  • Research Methods: Refers to the concrete behavior, instruments, computational scripts, and statistical procedures utilized to conduct research operations (e.g. administering a questionnaire, executing a \(t\)-test, or training a gradient boosted classifier).
  • Research Methodology: Refers to the overarching scientific philosophy, theoretical framework, and logical strategy that justifies why specific methods were selected, why alternative approaches were rejected, and how the overall study design guarantees valid inference.

1.2.1 Architectural Analogy:

  • Research Methods represent the physical construction tools (hammer, saw, level).
  • Research Methodology represents the structural engineering blueprint detailing why specific materials and structural loads were selected to guarantee building integrity.
Dimension Research Methods Research Methodology
Definition Specific tools, operations, and analytical techniques. The systematic, scientific framework justifying the study design.
Scope Narrow: targeted analytical execution (e.g. running an R function). Broad: encompasses research design, sampling philosophy, ethics, and validity.
Primary Goal To collect empirical data and compute numerical metrics. To establish the validity, reliability, and appropriateness of the overall approach.
Relation Methods are a constituent component of methodology. Methodology is the overarching framework hosting the methods.

1.3 The 11 Sequential Steps in the Research Process

Scientific inquiry follows an 11-step sequential process. Each stage logically depends upon the completion and rigor of the preceding step (Kothari and Garg (2019)):

Phase Sequential Steps Core Focus & Rationale
Phase I: Conceptualization Step 1: Formulate Research Problem
Step 2: Review Literature
Step 3: Develop Working Hypotheses		 Define operational variables, identify knowledge gaps, and formulate testable \(H_0\) vs \(H_1\).
Phase II: Design & Sampling Step 4: Prepare Research Design
Step 5: Determine Sample Design
Step 6: Collect Data		 Select observational vs experimental framework, calculate sample size \(n\), and execute data collection.
Phase III: Analysis & Reporting Step 7: Execute Project Quality Control
Step 8: Analyze Data
Step 9: Test Statistical Hypotheses
Step 10: Generalize & Interpret
Step 11: Prepare Research Manuscript		 Clean data, run inferential tests at \(\alpha=0.05\), evaluate validity, and write formal thesis/journal paper.

1.3.1 Logical Rationale for Sequential Progression:

  1. Formulating the Research Problem: Define the precise research question, target population, exposures, and primary outcomes. Rationale: Study design cannot proceed without clear operational definitions of variables.
  2. Extensive Literature Review: Synthesize existing peer-reviewed literature (Booth et al. (2016)). Rationale: Identifies the current state of knowledge and isolates the research gap, preventing redundant investigation.
  3. Developing Working Hypotheses: Formulate explicit testable statistical propositions:
    • Null Hypothesis (\(H_0\)): Statement of no difference, effect, or association.
    • Alternative Hypothesis (\(H_1\)): Statement of expected directional or non-directional effect.
  4. Preparing Research Design: Establish the overall structural framework (Observational vs. Experimental).
  5. Determining Sample Design: Establish the target population, sampling frame, sampling method, and required sample size (\(n\)).
  6. Executing Data Collection: Administer data collection instruments, extract electronic health records, or stream sensor metrics.
  7. Execution of the Project: Implement quality control, conduct pilot testing, and execute non-response management.
  8. Analysis of Data: Clean, transform, and structure data to compute descriptive and inferential statistics.
  9. Hypothesis Testing: Apply appropriate inferential statistical tests (\(t\)-tests, Chi-square, ANOVA, Cox regression) to evaluate \(H_0\) at significance level \(\alpha = 0.05\).
  10. Generalizations and Interpretation: Synthesize empirical findings, evaluate internal and external validity, and assess potential unmeasured confounding.
  11. Preparation of the Report / Manuscript: Document methodology, results, limitations, and conclusions in standard academic formats (Protocol, Thesis, or Peer-Reviewed Journal Article).

1.4 Research Ethics and Scientific Dissemination

Research ethics governs the moral standards and rules of conduct required during scientific investigation (Creswell and Creswell (2018)).

1.4.1 The Belmont Report Principles

  1. Respect for Persons (Autonomy): Requires that individuals be treated as autonomous agents capable of self-determination. Participants must provide voluntary Informed Consent after being fully apprised of study procedures, risks, and benefits. Vulnerable populations require enhanced safeguards.
  2. Beneficence: Obligations to maximize potential benefits to subjects and society while minimizing potential physical, psychological, or financial harms.
  3. Justice: Mandates equitable selection of research subjects such that the burdens and benefits of research are distributed fairly across societal groups.

1.4.2 Institutional Review Board (IRB) Review Categories

Before initiating human subjects research, protocols must undergo IRB evaluation:

  • Exempt Review: Research involving minimal risk and anonymized secondary datasets.
  • Expedited Review: Research involving minimal risk and non-invasive routine procedures.
  • Full Board Review: Research involving greater than minimal risk, novel interventions, or vulnerable subjects.

1.4.3 Scientific Misconduct and Academic Integrity

  • Fabrication: Recording or reporting fictitious data or results.
  • Falsification: Manipulating research materials, equipment, or processes, or omitting data such that the research is not accurately represented.
  • Plagiarism: Appropriating another author’s ideas, text, or results without proper attribution.

1.5 Literature Search, Referencing, and Citation Systems

1.5.1 Systematic Search Methodology and Literature Matrix

A literature search must be comprehensive and reproducible (Booth et al. (2016)):

  • Boolean Logic: Utilize operators AND (intersection), OR (union), and NOT (exclusion).
  • Core Bibliographic Databases: PubMed / MEDLINE (Biomedical), IEEE Xplore (Computer Science), arXiv (Preprints in ML/Statistics), Scopus / Web of Science (Multidisciplinary).

1.5.1.1 The Literature Synthesis Matrix:

A structured matrix utilized to compare prior empirical studies:

Author & Year Study Design Sample Size (\(n\)) Primary Methodology Key Empirical Findings Identified Limitation / Gap
Smith et al. (2021) Prospective Cohort \(n = 1,200\) Cox Proportional Hazards aHR = 1.45 (\(p < 0.01\)) Unadjusted missing EHR data
Patel et al. (2023) Double-Blind RCT \(n = 350\) Intention-to-Treat ANOVA Reduced recovery by 3.2 days Single-center design

1.5.2 Citation Systems

  • APA 7th Edition: Author-Date format widely utilized in social and data sciences. In-text: (Kothari & Garg, 2019).
  • Vancouver Style: Numbered sequence standard in medical literature. In-text: Evidence indicates elevated readmission [1].
  • BibTeX: Plain-text reference format utilized in LaTeX and R Markdown (@article{...}).

1.6 Formats: SPIRIT Protocol, Academic Thesis, and Journal Manuscript

Format Primary Purpose Timing relative to Data Collection Key Structural Feature
Research Protocol Formal pre-registration of study design & SAP Written & registered BEFORE data collection Contains NO Results section
Academic Thesis Multi-chapter university degree monograph Written AFTER project completion Follows 5-chapter structure or cumulative paper format
Journal Manuscript Peer-reviewed dissemination article Written AFTER analysis completion Follows IMRaD & reporting guidelines (CONSORT/STROBE/PRISMA)

1.6.1 Research Protocol (SPIRIT Statement Standard)

A research protocol is a formal prospective document prepared and registered prior to data collection.

  • Structural Feature: A protocol contains NO Results section because data collection has not occurred.
  • Components: Background, Rationale, Objectives, Inclusion/Exclusion criteria, Sample size power calculation, Statistical Analysis Plan (SAP), and IRB ethics approval.

1.6.2 Academic Thesis Architecture (5 Archetypes)

An academic thesis is a multi-chapter university monograph documenting degree candidate research:

  1. 🎓 Traditional Master’s Thesis: Standard 5-chapter monograph (Introduction, Literature Review, Methodology, Results, Discussion).
  2. 🏛️ PhD Doctoral Dissertation: Advanced dissertation featuring novel mathematical proofs and simulation frameworks.
  3. 📄 Article-Based Thesis: Monograph compiling published peer-reviewed manuscripts bound by an integrative synthesis.
  4. 🗣️ Qualitative Thesis: Replaces quantitative tables with narrative themes, participant transcript quotes, and reflexivity.
  5. 💻 Data Science Thesis: Focuses on algorithmic developments, Monte Carlo simulations, and open-source software packages.

1.6.3 Journal Manuscripts (IMRaD & Reporting Guidelines)

Peer-reviewed journal articles condense empirical studies into 8-12 page manuscripts structured according to IMRaD:

  • Introduction: Problem formulation, background, gap, and hypotheses.
  • Methods: Detailed protocol permitting exact independent replication.
  • Results: Empirical findings presenting summary tables, figures, effect sizes, and \(p\)-values.
  • Discussion: Contextual interpretation of findings, comparison with prior literature, and study limitations.

1.6.3.1 International Reporting Guidelines:

  • CONSORT 2010: Reporting checklist for Randomized Controlled Trials (includes 4-phase participant flow diagram).
  • STROBE Statement: Reporting checklist for Observational Studies (Cohort, Case-Control, Cross-Sectional).
  • PRISMA 2020: Reporting checklist for Systematic Reviews and Meta-Analyses.

1.7 Research Design: Taxonomy, Validity, and Case Studies

Research design is the overarching blueprint for empirical data collection and statistical analysis (Kothari and Garg (2019)).

1.7.1 Key Structural Elements

  1. Variables:
    • Dependent Variable (\(Y\)): Primary outcome measure under investigation.
    • Independent Variable (\(X\)): Predictor or exposure variable manipulated or observed.
    • Confounding Variable: An unmeasured or unadjusted extraneous factor associated with both exposure \(X\) and outcome \(Y\).
  2. Experimental Controls:
    • Experimental Group: Subjects receiving the active intervention.
    • Control Group: Subjects receiving placebo or standard-of-care baseline comparison.
  3. Blinding Protocols:
    • Single-Blind: Subjects are unaware of arm assignment.
    • Double-Blind: Neither subjects nor outcome assessors are aware of arm assignment (eliminates expectations bias).

1.7.2 Research Design Taxonomy

Design Type Temporal Orientation Primary Outcome Metric Major Advantage Major Limitation Best Application
Cross-Sectional Single point in time Point Prevalence Rapid execution, cost-effective Cannot establish temporal causality Population prevalence surveys
Case-Control Retrospective Odds Ratio (\(OR = \frac{ad}{bc}\)) Efficient for rare outcomes Vulnerable to recall bias Investigating rare clinical conditions
Prospective Cohort Longitudinal / Prospective Relative Risk (\(RR\)), Hazard Ratio (\(HR\)) Establishes temporal sequence High cost, attrition bias Tracking risk factor incidence
Randomized Trial (RCT) Experimental / Prospective Risk Ratio, Mean Difference Gold standard for causal inference High cost, strict inclusion limits Evaluating therapeutic efficacy

1.7.3 Experimental Design Classifications

  • Completely Randomized Design (CRD): Treatments assigned completely at random across homogeneous units.
  • Randomized Block Design (RBD): Units grouped into homogeneous blocks based on a nuisance variable (e.g. age strata) prior to randomization.
  • Latin Square Design (LSD): Simultaneously controls for two nuisance sources of variation using a square matrix layout.

1.7.4 Threats to Validity

  • Internal Validity: The degree to which a study cleanly isolates a true causal relationship without confounding (threatened by selection bias, maturation, attrition, and instrumentation changes).
  • External Validity: The degree to which empirical findings generalize to broader target populations (threatened by restrictive eligibility criteria and Hawthorne effects).

1.8 Sampling Design and Sample Size Determination

Sampling design is a definite, systematic plan for obtaining a representative sample from a target population (Kothari and Garg (2019)).

1.8.1 Core Sampling Terminology

  • Target Population: The complete collection of units about which theoretical inferences are drawn.
  • Sampling Frame: The physical register or list of population units available for selection.
  • Sampling Error: The random statistical variation resulting from estimating a population parameter using a sample rather than a census (\(SE(\bar{x}) \propto 1/\sqrt{n}\)).

1.8.2 Probability Sampling Methods & Mathematical Derivations

In probability sampling, every unit \(i\) in population \(N\) possesses a known, non-zero inclusion probability \(\pi_i > 0\).

1.8.2.1 Simple Random Sampling (SRS)

Every subset of \(n\) units has equal selection probability \(P = 1/N\).

  • Proof of Sample Mean Unbiasedness: \(E[\bar{y}] = \bar{Y}\).
  • Variance of Estimator (SRSWOR): \(Var(\bar{y}) = \frac{S^2}{n} (1 - f)\), where \(f = n/N\) is the sampling fraction and \((1-f)\) is the Finite Population Correction (FPC).

1.8.2.2 Stratified Random Sampling & Optimum Allocations

Divide population \(N\) into non-overlapping strata \(N_1, \dots, N_K\). Draw an SRS from each stratum.

  • Proportional Allocation: \(n_h = n \left(\frac{N_h}{N}\right)\).

  • Neyman Optimum Allocation (Variance Minimization): \[n_h = n \cdot \frac{N_h S_h}{\sum_{k=1}^K N_k S_k}\]

  • Cost-Constrained Optimum Allocation (Minimizing Variance under Budget \(C = c_0 + \sum c_h n_h\)): \[n_h = n \cdot \frac{\frac{N_h S_h}{\sqrt{c_h}}}{\sum_{k=1}^K \frac{N_k S_k}{\sqrt{c_k}}}\]

1.8.2.2.1 📝 Worked Numerical Problem 1: Cost-Constrained Allocation

A survey samples \(n = 500\) subjects across 2 geographic strata:

  • Stratum 1: Population \(N_1 = 10,000\), standard deviation \(S_1 = 20\), cost per unit \(c_1 = \$4\).
  • Stratum 2: Population \(N_2 = 5,000\), standard deviation \(S_2 = 30\), cost per unit \(c_2 = \$9\).

Derivation & Solution:

  1. Evaluate \(\frac{N_h S_h}{\sqrt{c_h}}\):
    • Stratum 1: \(\frac{10,000 \times 20}{\sqrt{4}} = \frac{200,000}{2} = 100,000\).
    • Stratum 2: \(\frac{5,000 \times 30}{\sqrt{9}} = \frac{150,000}{3} = 50,000\).
    • Sum \(\sum \frac{N_k S_k}{\sqrt{c_k}} = 100,000 + 50,000 = 150,000\).
  2. Compute optimum stratum sample sizes \(n_h\):
    • \(n_1 = 500 \times \frac{100,000}{150,000} = \mathbf{333 \text{ units}}\).
    • \(n_2 = 500 \times \frac{50,000}{150,000} = \mathbf{167 \text{ units}}\).

1.8.2.3 Systematic Sampling

Select every \(k\)-th element following a random start \(r \in \{1, \dots, k\}\), where \(k = \lfloor N/n \rfloor\).

  • Advantage: High operational efficiency in field settings.
  • Limitation: Highly vulnerable to bias when population frames exhibit periodic cycles.

1.8.2.4 Cluster & Multi-Stage Sampling

Divide population into \(M\) primary clusters. Sample \(m\) clusters and survey all secondary units.

  • Design Effect (\(DEFF\)): Quantifies variance inflation due to intra-cluster correlation \(\rho\): \[DEFF = \frac{Var_{cluster}}{Var_{SRS}} = 1 + (\bar{M} - 1)\rho\]
    Higher intra-cluster correlation (\(\rho > 0\)) inflates variance, requiring sample size adjustment (\(n_{eff} = n / DEFF\)).

1.8.3 Non-Probability Sampling Methods

Non-random selection strategies useful for exploratory research, but invalid for statistical population generalization:

  • Convenience Sampling: Selection based on accessibility.
  • Purposive Sampling: Expert judgemental selection.
  • Quota Sampling: Non-random demographic quota filling.
  • Snowball Sampling: Chain-referral sampling utilized for hard-to-reach hidden populations.

1.8.4 Sample Size Determination Formulas

1.8.4.1 Estimating a Single Population Proportion \(p\):

\[n = \frac{Z_{1-\alpha/2}^2 \cdot p(1 - p)}{d^2}\]

1.8.4.2 Estimating a Single Population Mean \(\mu\):

\[n = \frac{Z_{1-\alpha/2}^2 \cdot \sigma^2}{d^2}\]

1.8.4.2.1 📝 Worked Numerical Problem 2: Sample Size Calculation

Estimate mean fasting blood glucose within a margin of error \(d = 3 \text{ mg/dL}\) at \(95\%\) confidence (\(Z = 1.96\)), given population standard deviation \(\sigma = 25 \text{ mg/dL}\).

Solution: \[n = \frac{(1.96)^2 \times 25^2}{3^2} = \frac{3.8416 \times 625}{9} = \frac{2401}{9} = 266.77 \implies \mathbf{267 \text{ subjects}}\]

1.8.5 🎓 Unit 1 Key Takeaways

  • Methods vs. Methodology: Methods are concrete operations; methodology is the theoretical justification of validity.
  • 11-Step Process: Sequential framework from problem formulation through literature search to report preparation.
  • Reporting Standards: SPIRIT for protocols, CONSORT for RCTs, STROBE for observational studies, PRISMA for reviews.
  • Research Design Taxonomy: Observational (Cross-sectional, Case-Control \(OR\), Cohort \(HR\)) vs Experimental (RCTs, RBD, LSD).
  • Sampling Design: SRS, Stratified (Neyman & Cost-constrained allocation), Systematic, Cluster (\(DEFF = 1 + (\bar{M}-1)\rho\)).

References

Booth, Wayne C, Gregory G Colomb, Joseph M Williams, Joseph Bizup, and William T Fitzgerald. 2016. The Craft of Research. 4th ed. University of Chicago Press.
Creswell, John W, and J David Creswell. 2018. Research Design: Qualitative, Quantitative, and Mixed Methods Approaches. 5th ed. SAGE Publications.
Kothari, C. R., and Gaurav Garg. 2019. Research Methodology: Methods and Techniques. Fourth. New Age International (P) Limited.

Statistical Research Methodology (DDS 6106) • Author: Dr. Shrikrishna Bhat K • Department of Applied Statistics and Data Science, PSPH, MAHE