Thinking Like a Researcher – Chapter 2 Study Notes

Retraining the Mind to “Think Like a Researcher”

  • Core skill: shuttling mentally between
    • Empirical plane (direct observations/data) and
    • Theoretical plane (abstract concepts, laws, generalizations).
  • Requires visualizing abstractions, spotting hidden patterns, synthesizing them into generalizable principles.
  • Not formally taught; most common deficit among beginning Ph.D. students.
  • Key intellectual tools (examined throughout the chapter):
    • Unit of analysis
    • Concepts / constructs / variables
    • Hypotheses & propositions
    • Operationalization
    • Theories & models
    • Induction vs. deduction

Unit of Analysis (UoA)

  • Definition: The person, collective, or object that is the primary target of investigation.
  • Common UoA categories & examples:
    • Individuals: studying shopping behavior, learning outcomes, technology attitudes.
    • Groups: street gangs, organizational teams.
    • Organizations/firms: profitability, executive decision quality.
    • Countries/nations: cultural differences.
    • Technologies/objects: web pages’ attractiveness.
    • Dyads: knowledge‐transfer between two firms.
  • Complexities & nuanced cases:
    • Neighborhood crime study → UoA = neighborhood, not the criminal.
    • Comparing crime types across neighborhoods → UoA = crime.
    • Studying criminals’ motives → UoA = individual criminal.
    • Innovation success of products vs. innovative capability of firms → UoA switches between innovation (object) and organization.
  • Implications for data collection:
    • Data must come from the UoA.
    • UoA = web page → scrape page attributes; do not survey users.
    • UoA = organization → gather firm‐level metrics (size, revenue, hierarchy, absorptive capacityabsorptive\ capacity) from reports or CEO surveys.
    • Some ostensibly individual metrics (e.g., CEO pay) may be organizational variables (each firm has a single CEO compensation value).
  • Aggregation possibility:
    • Collect lower‐level data and statistically aggregate upward; e.g., survey team members → average into team‐level cohesion/conflict scores.

Concepts, Constructs & Variables

  • Goal of explanatory research: generate explanations via generalizable characteristics.
  • Terminology hierarchy:
    • Object: tangible entity (person, firm, car) → not a concept.
    • Concept: characteristic/behavior of object (attitude, innovation capacity, car weight).
    • Everyday language, borrowed metaphors (gravity in shopping), or newly coined (technostress).
    • Vary in abstraction: weight (concrete) vs. personality (abstract).
    • Construct: an abstract concept specifically created/selected to explain a phenomenon.
    • Unidimensional (weight) vs. multidimensional (communication skill → vocabulary, syntax, spelling).
    • For multidimensional cases: higher‐order term = construct; lower‐order terms = concepts.
  • Importance of precise definitions:
    • Ambiguous example: “income” → must clarify monthly/yearly, before/after tax, personal/family.
  • Two definition styles:
    • Dictionary/synonym (circular; inadequate).
    • Operational definition: spells out empirical measurement method (temperature in ∘C^{\circ}C, ∘F^{\circ}F, or K).
  • Variable:
    • Etymology: something that can vary.
    • In research: measurable proxy for an abstract construct.
    • Example: construct = intelligence; variable = IQ score (pattern‐matching test index).
    • Measurement quality: variable can be good/poor proxy depending on validity.
  • Two‐plane viewpoint (Fig. 2.1):
    • Theoretical plane → constructs.
    • Empirical plane → variables.
    • Research skill = moving between planes.
  • Variable roles:
    • Independent (predictor/cause)
    • Dependent (outcome/effect)
    • Mediating (transmits effect from IV to DV)
    • Moderating (changes strength/direction of IV–DV link)
    • Control (extraneous; held constant/statistically controlled)
  • Example network (Fig. 2.2):
    • Intelligence (IQIQ) → Academic Achievement (GPA).
    • Effort moderates the above effect.
    • Academic Achievement mediates path from Intelligence to Earning Potential.
    • Full set of linked constructs = nomological network.

Propositions & Hypotheses

  • Proposition: Declarative, theoretical statement of relationship between constructs. Example: “Higher intelligence increases academic achievement.”
    • Must be empirically testable.
    • Derived via logic (deduction) or observation (induction).
  • Hypothesis: Empirical statement linking measured variables, derived from proposition. Example: “Higher IQ scores lead to higher GPA.”
  • Strength spectrum:
    • Weak: “IQ related to GPA.” (no direction, causality)
    • Better: “IQ positively related to GPA.” (direction)
    • Strong: “IQ positively affects GPA.” (direction + causality + IV/DV clarity)
  • Criteria for scientific hypothesis:
    • Specifies IV and DV.
    • States directional/causal expectation.
    • Falsifiable (true/false via data).
    • Non‐conforming examples: vague statements like “students are intelligent.”

Theories

  • Definition: Systematically interrelated constructs & propositions designed to explain/predict a phenomenon within stated boundaries & assumptions.
  • Scale: broader, more complex and abstract than single propositions/hypotheses.
  • Misconception correction: theory ≠ speculation; practice ≠ opposite of theory.
    • Lewin: “Theory without practice is sterile; practice without theory is blind.”
  • Theories vary in quality; evaluated via explanatory power, parsimony, falsifiability (discussed Ch. 3).
  • Scientific progress = replacing poorer theories with better ones.

Models

  • Definition: Representation (often partial) of a system constructed to study or analyze that system.
  • Functional distinction: Theory explains; model represents/helps decision‐making.
  • Usage examples:
    • Marketing mix model → decide advertising spend.
    • Weather model → forecast weather from wind, temperature, humidity.
  • Model typologies:
    • By form: Mathematical, network, path.
    • By purpose: Descriptive (depict complexity), Predictive (forecast future; e.g., regression), Normative (guide action via norms).
    • By time: Static (snapshot) vs. Dynamic (evolution over time).

Induction & Deduction in Model/Theory Building

  • Deduction: Draw conclusion from premises via logic.
    • Bank enforces strict ethics (P1); Jamie works there (P2) → Jamie behaves ethically (conclusion). True if premises true.
  • Induction: Infer generalization/explanation from observed facts.
    • Heavy promotion, but no sales lift → hypothesize poor campaign execution. Competing explanations possible → conclusions tentative.
  • Relative strength:
    • Deductive conclusions stronger if premises correct.
    • Inductive conclusions inherently probabilistic.
  • Interplay in research (Fig. 2.3):
    1. Observe fact (induction triggers “why?”).
    2. Generate multiple tentative explanations (hypotheses).
    3. Apply deduction to narrow to most plausible explanation.
    4. Empirically test; loop back with new observations → refine theory/model.
  • Essential researcher skill: fluid movement between induction and deduction when extending/modifying theories or building better ones.