Quantitative Research Designs – Practical Research 2

Overview: Kinds of Quantitative Research Design

  • The lecture outlines five major quantitative research designs used in Practical Research 2:
    • Descriptive (Observational) Research Design
    • Correlational Research Design
    • Ex-Post Facto (Retrospective) Research Design
    • Quasi-Experimental Research Design
    • Experimental Research Design
  • Each design answers different research questions, entails distinct levels of control over variables, and carries unique ethical or logistical considerations.

Descriptive / Observational Research Design

  • Purpose: Systematically describe characteristics or patterns without manipulating variables.
  • Assignment 1 Context: “An observational study describing traffic patterns in a city.”
  • Core Variables Typically Measured in Traffic Studies:
    • Traffic Volume (Flow)
    • Definition: Number of vehicles passing a specific point or road segment within a set period (e.g., vehicles/hour).
    • Practical relevance: Determines infrastructure capacity needs and peak-hour congestion management.
    • Traffic Speed
    • Definition: Velocity at which vehicles travel; may reference average speed, 85th-percentile speed, or spot speed.
    • Significance: Serves as a safety indicator and informs speed-limit policies.
    • Traffic Density / Congestion
    • Definition: Number of vehicles occupying a defined road length at an instant (density) or the severity of flow impedance (congestion).
    • Real-world link: High density often correlates with accident risk and reduced economic productivity due to delays.
    • Travel Time / Delay
    • Definition: Time required for a vehicle to traverse a route or additional time incurred from congestion.
    • Usage: Key metric in level-of-service analysis and cost–benefit studies.
  • Ethical/Practical Implications:
    • Minimal risk to participants—data often collected passively via sensors or cameras.
    • Must address privacy when recording license plates or using GPS traces.

Correlational Research Design

  • Goal: Determine the degree and direction of association between two or more variables without implying causation.
  • Key Statistical Tools Introduced:
    • Pearson’s Correlation Coefficient (r)
    • Measures linear relationships between two continuous variables.
    • Formula: r=∑(X−Xˉ)(Y−Yˉ)∑(X−Xˉ)2 ∑(Y−Yˉ)2r = \frac{\sum (X - \bar{X})(Y - \bar{Y})}{\sqrt{\sum (X - \bar{X})^2\, \sum (Y - \bar{Y})^2}}
    • Range: −1≤r≤1-1 \le r \le 1 ( +1 perfect positive, 0 none, –1 perfect negative).
    • Spearman’s Rank-Order Correlation Coefficient (\rho)
    • Non-parametric metric for monotonic relationships between ordinal or ranked data.
    • Formula (when no tied ranks): ρ=1−6∑d<em>i2n(n2−1)\rho = 1 - \frac{6 \sum d<em>i^2}{n(n^2 - 1)} where d</em>id</em>i is rank difference per case.
    • Likert Scale
    • An ordinal scaling technique (commonly 5- or 7-point) where respondents indicate agreement/intensity (e.g., 1 = Strongly Disagree to 5 = Strongly Agree).
    • Facilitates conversion of attitudes into numeric form suitable for correlation.
  • Example 1 (Pearson): Study Hours vs. Exam Scores
    • Variables: Hours studied (continuous) & exam score (continuous).
    • Hypothetical outcome: r=0.75r = 0.75 ⇒ strong positive linear relation (more study hours ➜ higher scores).
    • Practical implication: Though correlated, unmeasured factors (prior knowledge, motivation) could confound causation.
  • Example 2 (Spearman): Judges’ Rankings in a Cooking Competition
    • Variables: Rank given by Judge A vs. rank by Judge B on 10 dishes.
    • If ρ=0.85\rho = 0.85 ⇒ strong agreement; if low/negative ⇒ disagreement or opposite tastes.
  • Ethical/Philosophical Notes:
    • Correlation ≠ causation—misinterpretation can lead to faulty policy decisions.
    • Researchers must transparently report other plausible explanations.

Ex-Post Facto Research Design (“After-the-Fact” Studies)

  • Definition: Investigates possible cause-and-effect relationships by examining existing conditions/events; researcher does NOT manipulate the independent variable.
  • Analogy: Detective work—effect ("crime") has occurred; researcher looks backward for possible causes.
  • Typical Steps:
    1. Identify effect/outcome that already exists.
    2. Retrieve archival data or survey participants about prior exposure.
    3. Compare groups differing in past conditions.
  • Example 1: Childhood exposure to environmental toxins vs. later-life health issues.
  • Example 2: Private vs. public high-school attendance impacting college achievement.
  • Strengths & Limitations:
    • Can explore ethically sensitive topics where manipulation isn’t possible.
    • Susceptible to confounds; cannot establish definitive causation.
    • Requires rigorous control of extraneous variables through matching or statistical adjustments (e.g., ANCOVA).

Quasi-Experimental Research Design

  • Purpose: Estimate causal effects when random assignment is impractical or unethical.
  • Hallmarks:
    • Presence of an intervention (independent variable) and outcome (dependent variable).
    • Comparison/control group exists, but participants are in pre-existing groups (no randomization).
    • Often includes pre-test and post-test to gauge change over time.
  • Example: Activity-Based Math Curriculum Evaluation
    • Setting: School A (new curriculum) vs. School B (traditional); schools matched on size & demographics.
    • Independent Variable: Curriculum type.
    • Dependent Variables: Student engagement (observations, surveys) & math test scores.
    • Strengthening design: Collect baseline data to control initial differences; may apply statistical controls (propensity scores).
  • Ethical & Practical Considerations:
    • More feasible than true experiments in educational or medical contexts.
    • Must discuss threats to internal validity: selection bias, maturation, history, instrumentation.

Experimental Research Design (True Experiment)

  • Core Features:
    • Direct manipulation of one or more independent variables.
    • Random assignment to experimental vs. control groups.
    • Control of extraneous variables, often via standardized procedures or laboratory settings.
  • Primary Goal: Establish strong evidence for cause-and-effect.
  • Example: Online Interactive Module vs. Traditional Textbook
    • Participants: 100 Grade 10 students randomly allocated (50/50 split).
    • Experimental Group: Uses online interactive learning module for one week.
    • Control Group: Uses traditional textbook for same content & time.
    • Outcomes: Post-intervention test on scientific concepts.
    • Randomization ensures group equivalence, enhancing internal validity.
  • Advantages:
    • Highest level of causal inference among quantitative designs.
    • Results can inform instructional policy or product adoption.
  • Ethical Notes:
    • Must secure informed consent, especially for educational experiments.
    • Equitable access: After study, control group may be offered intervention to avoid unfair deprivation.

Comparative Reflection & Decision Tree

  • If you want to describe or profile a phenomenon ⟹ Descriptive design.
  • If you want to estimate relationships (no causation) ⟹ Correlational design.
  • If the cause happened in the past & you cannot manipulate ⟹ Ex-Post Facto.
  • If you can manipulate but cannot randomize ⟹ Quasi-Experimental.
  • If you can manipulate and randomize ⟹ Experimental.

Cross-Design Ethical & Practical Implications

  • Data Privacy: Traffic cameras, student records, or health data all require confidentiality safeguards.
  • Informed Consent: Mandatory for any study collecting identifiable personal data or testing interventions.
  • Generalizability vs. Control Trade-off:
    • Descriptive & correlational designs often capture real-world settings (high external validity).
    • Experimental designs maximize internal validity but may suffer from artificiality.
  • Sequential Research Path:
    1. Start descriptive ➜ identify patterns.
    2. Move to correlational ➜ find associations.
    3. Proceed to quasi-experimental ➜ test causality where randomization is impractical.
    4. Culminate in experimental ➜ confirm cause-effect under controlled conditions.

Key Numerical/Statistical Concepts to Remember

  • Pearson’s rr: linear correlation for interval/ratio data.
  • Spearman’s ρ\rho: ranked/ordinal data correlation, resilient to non-normality.
  • Likert Data: Ordinal; treat cautiously—means & Pearson’s rr are common but debated.
  • Pre-test/Post-test Difference Scores: Δ=Post−Pre\Delta = \text{Post} - \text{Pre} to gauge intervention impact.
  • Effect Size Benchmarks (Cohen): ∣r∣≈0.10|r| \approx 0.10 (small), 0.300.30 (medium), 0.500.50 (large).

Study Tips for the Exam

  • Memorize definitions & be able to match designs to real-life scenarios.
  • Practice computing and interpreting Pearson’s rr and Spearman’s ρ\rho.
  • Be ready to critique a study’s design—identify design type, strengths, and threats to validity.
  • Link ethical considerations to each design (e.g., why randomizing smoking status is unethical).
  • Use the detective metaphor to quickly recall ex-post facto logic.