Bivariate Data Analysis and Quantitative Research Design

Research Design and Structural Alignment in Quantitative Analysis

• Alignment of Core Assessment Elements: Successful quantitative analysis relies on establishing a coherent research design and data narrative rather than performing complex statistical procedures. Technical execution in basic bivariate analysis is straightforward, whereas structural framing requires tight integration across several components:
  • Problem Statement: Defines the central context, scope, and significance of the study.
  • Literature Review: Synthesizes prior theoretical frameworks and empirical findings to contextualize the core topic.
  • Research Question: Formulated directly from the literature review to target specific knowledge gaps.
  • Analysis Plan: Dictates the explicit statistical methods and data breakdowns required to address the research question.
  • Findings and Discussion: Presents analytical results to directly answer the research question while validating, challenging, or extending prior literature.
• Precision and Analytical Focus: Research design must maintain a singular, consistent analytical angle across all sections, ensuring the proposed narrative aligns perfectly with the chosen data subsets.

Bivariate Analysis and Variable Identification

• Definition of Bivariate Analysis: The simultaneous statistical analysis of two variables to assess whether an association, correlation, or systematic relationship exists between them.
• Independent Variables ($IV$):
  • Conceptual Definition: The predictor, explanatory, or cause variable. It is hypothesized to influence or drive changes in another variable.
  • Experimental Context: The variable systematically manipulated by the researcher to observe its precise impact on an outcome.
  • Plant Exposure Example: Investigating whether sunlight exposure impacts plant growth rate.
  • Experimental Setup: Identical plants receive uniform daily water, identical soil quality, and identical fertilizer, while daily sunlight exposure is varied across experimental groups ($6\, \text{hours}$, $5\, \text{hours}$, $4\, \text{hours}$, $3\, \text{hours}$, $2\, \text{hours}$, $1\, \text{hour}$, and $0\, \text{hours}$).
  • Variable Role: Daily sunlight exposure duration functions as the independent variable ($IV$).
  • Socioeconomic Example: Evaluating how educational attainment influences total personal income; education level serves as the independent variable ($IV$).
• Dependent Variables ($DV$):
  • Conceptual Definition: The outcome or effect variable. It changes in direct response to alterations in the independent variable.
  • Plant Exposure Example: Daily growth measured in centimeters over a timeframe of $1\, \text{year}$ serves as the dependent variable ($DV$).
  • Socioeconomic Example: Personal income level serves as the dependent variable ($DV$).
• Covariates and Confounding Factors:
  • Conceptual Definition: Additional, extraneous variables that influence the dependent variable (or correlate with the independent variable) but are not the primary focus of the research question.
  • Socioeconomic Covariates: Age, work experience, geographic region, and industry sector.
  • Botanical Covariates: Soil quality, fertilizer type and amount, ambient temperature, and water volume.
  • Complex Social Phenomena: Social processes involve multiple interacting factors. Researchers must competently identify relevant covariates and explicitly report unobserved confounding factors as study limitations.

Study Designs, Causality, and Spurious Relationships

• Methodological Hierarchy for Establishing Causality:
  • Experimental Designs: Recognized as the definitive gold standard for establishing true causal relationships ($IV \rightarrow DV$) due to direct researcher control over variable manipulation and assignment.
  • Quasi-Experimental and Longitudinal Designs: Can approach causal inference more effectively than cross-sectional studies by capturing temporal order across time.
  • Cross-Sectional Designs: Represent a single observational snapshot of a population at one specific point in time. They detect associations or correlations but cannot prove causality or temporal direction.
• Bidirectionality in Cross-Sectional Data:
  • In cross-sectional studies examining education and income, directionality cannot be isolated empirically:
  • Higher family income may enable individuals to attain higher education degrees.
  • Higher education degrees may enable individuals to secure higher-paying positions.
  • A continuous feedback loop may exist where income and education repeatedly reinforce one another.
• Spurious Relationships:
  • Definition: An observed statistical association between two variables that is misleading or false, caused entirely by an unobserved third variable (a confounding factor) that independently drives both variables.
  • Example 1: Ice Cream Consumption and Drowning Incidents:
  • Empirical Observation: Daily ice cream sales correlate positively with drowning deaths in waterways.
  • Real Explanation: The relationship is spurious. Ambient air temperature (summer season) drives both increased ice cream consumption and higher rates of recreational swimming, leading to elevated drowning risks.
  • Example 2: Shoe Size and Reading Competency:
  • Empirical Observation: Larger shoe size correlates positively with higher reading scores among children.
  • Real Explanation: The relationship is spurious. Child age drives both physical foot size growth and cognitive reading development.
• Analytical Rigor: Researchers must evaluate empirical correlations critically, consult prior literature to identify unmeasured confounding variables, and explicitly state risks of spuriousness in study limitations.

Strategies for Variable Designation in Cross-Sectional Research

• Justifying $IV$ and $DV$ Roles Without Direct Causal Manipulation:
  • Prior Empirical Research and Theoretical Frameworks: Use established theoretical models (e.g., Social Identity Theory or Threat Theory positing that perceived identity threats cause prejudice) to justify variable placement.
  • Temporal Logic: Determine chronological sequencing when survey questions capture past events.
  • Example: Experiencing or witnessing a vilification incident in the preceding $12\, \text{months}$ logically precedes a respondent's baseline knowledge of anti-vilification laws measured prior to starting an educational unit.
  • Demographic Anchors: Fixed demographic characteristics that cannot logically be altered by non-demographic outcome variables.
  • Example: Age predicting legal awareness; awareness levels cannot alter a respondent's chronological age.

Cross-Tabulation (Crosstab) Analysis

• Definition and Structure: A foundational bivariate analytical technique for categorical data, displaying the joint distribution of two variables in a matrix format containing absolute frequencies ($n$) and relative percentages.
• Structural Components of a Crosstab Matrix:
  • Row Variables: Categories arranged horizontally across rows.
  • Column Variables: Categories arranged vertically down columns.
  • Absolute Frequencies ($n$): The raw numerical count of observations meeting specific row-column criteria.
  • Row Percentages: Calculated horizontally across each row, summing to $100\%$ at the rightmost margin. Indicates how observations within a specific row category are distributed across column categories.
  • Column Percentages: Calculated vertically down each column, summing to $100\%$ at the bottom margin. Indicates how observations within a specific column category are distributed across row categories.
• Empirical Interpretation Example (Study Mode Preference by Age Group):
  • Row Categories: Age Over $30$ ($n = 200$ total), Age Under $30$ ($n = 200$ total).
  • Column Categories: Individual Study Mode, Group Study Mode.
  • Data Matrix (Row Percentages):
  • Age Over $30$: $120$ prefer individual study ($60\%$), $80$ prefer group study ($40\%$).
  • Age Under $30$: $90$ prefer individual study ($45\%$), $110$ prefer group study ($55\%$).
  • Analytical Narrative: Age functions as a demographic anchor ($IV$). A distinct pattern reveals that while the majority ($60\%$) of respondents over $30$ prefer individual study, the majority ($55\%$) of respondents under $30$ prefer group study. If no relationship existed, percentage distributions across age brackets would be uniform.

Aligning Percentage Selection with Research Questions

• Determining Percentage Types Based on Research Objectives:
  • Question Type A: "Does age distribution differ between women and men?"
  • Percentage Requirement: Row Percentages (when gender is positioned in rows and age brackets in columns). Evaluates the proportional distribution of age within the female subgroup and within the male subgroup.
  • Question Type B: "Does gender composition differ across age groups?"
  • Percentage Requirement: Column Percentages (when age brackets are positioned in columns). Evaluates the proportional split of gender within each distinct age bracket.
• Empirical Statement Evaluation Matrix (Rows = Women, Men; Columns = Age Groups $18\text{--}29$, $30\text{--}49$, $50+$; Row Percentages Used):
  • Provided Row Data:
  • Women Subgroup: $40\%$ aged $18\text{--}29$, $35\%$ aged $30\text{--}49$, $25\%$ aged $50+$.
  • Men Subgroup: $25\%$ aged $18\text{--}29$, $35\%$ aged $30\text{--}49$, $40\%$ aged $50+$.
  • Statement Correctness Analysis:
  • Statement 1: "$40\%$ of women are aged $18\text{ to }29$." -> Correct (accurately reflects the row percentage for women).
  • Statement 2: "$40\%$ of people aged $18\text{ to }29$ are women." -> Incorrect (requires column percentages to calculate gender proportions within that specific age bracket).
  • Statement 3: "Women are more likely than men to be aged $18\text{ to }29$." -> Correct (directly compares the $40\%$ female rate against the $25\%$ male rate within that bracket).
  • Statement 4: "Among people aged $50$, $40\%$ are men." -> Incorrect (requires column percentages to calculate gender proportions within the $50+$ cohort).

Textual Reporting and Data Visualization Guidelines

• Narrative Guidelines for Reporting Crosstabs:
  • Avoid Total Numerical Repetition: Do not restate every single figure from a crosstab inside the body text.
  • State Primary Patterns: Use text to articulate the dominant narrative, trend, or directional pattern revealed by the data.
  • Selective Number Integration: Embed $1$ or $2$ specific, impactful figures to concretely substantiate the primary finding.
  • Contextual Integration: Connect observed empirical trends directly to theoretical predictions or past study outcomes.
  • Exemplar Text (Online Hate Exposure by Age Brackets):
  • Pattern Statement: Exposure to online hate decreases monotonically with age.
  • Substantive Evidence: Frequent exposure is reported by $45\%$ of respondents aged $18\text{ to }29$, $25\%$ of those aged $30\text{--}49$, and $10\%$ of those aged $50+$.
  • Narrative Context: This empirical pattern aligns with theoretical expectations that younger populations spend greater cumulative time on digital platforms, increasing opportunities for exposure.
• Visualization Formats for Bivariate Categorical Data:
  • Clustered Column / Bar Charts: Positions discrete category bars side by side. Ideal for directly comparing subgroups across static categories (e.g., comparing male vs. female rates across age cohorts).
  • Stacked Column / Bar Charts: Displays each total group category as a single bar partitioned into colored sub-segments. Ideal for visualizing proportional composition within whole categories at a single glance.

Excel Workflow for Pivot Table Analysis and Charting

• Codebook Variables Analyzed:
  • Variable $Q2$: Baseline awareness of statutory changes to Victoria's anti-vilification laws prior to unit enrollment ("Aware" vs. "Not Aware").
  • Variable $Q17$: Country of birth ("Were you born in Australia?" -> "Yes" vs. "No").
  • Variable $Q16$: Student enrollment classification ("Domestic Student" vs. "International Student").
• Step-by-Step Execution in Excel:
  1. Select Complete Dataset: Press Ctrl + A within the data worksheet.
  2. Insert Pivot Table: Navigate to Insert -> PivotTable -> Select New Worksheet.
  3. Assign Pivot Table Fields:
  • Drag Independent Variable ($IV$) into Columns (e.g., Variable $Q17$: Country of birth).
  • Drag Dependent Variable ($DV$) into Rows (e.g., Variable $Q2$: Awareness of anti-vilification laws).
  • Drag Dependent Variable ($DV$) into Values (set field summarize by to Count of Q2).
  1. Calculate Column Percentages: Right-click any value cell inside the table -> Select Show Values As -> Choose % of Column Total.
  2. Format Decimal Precision: Right-click value -> Select Number Format -> Adjust to $0$ decimal places.
• Empirical Results Breakdown 1 ($Q17$ Country of Birth vs. $Q2$ Law Awareness):
  • Raw Frequencies ($n$):
  • Overseas-Born Cohort ($n = 15$ total): $14$ Not Aware, $1$ Aware, $1$ blank non-response.
  • Australian-Born Cohort ($n = 37$ total): $24$ Not Aware, $13$ Aware.
  • Column Percentages:
  • Overseas-Born Cohort ($n = 15$): $93\%$ Not Aware, $7\%$ Aware.
  • Australian-Born Cohort ($n = 37$): $65\%$ Not Aware, $35\%$ Aware.
  • Empirical Finding: Australian-born students exhibit substantially higher pre-existing awareness of anti-vilification laws ($35\%$) compared to overseas-born students ($7\%$).
  • Confounding Variable Evaluation ($Q16$ Student Status):
  • Hypothesis: Student enrollment status ($Q16$ - Domestic vs. International) explains the relationship observed in $Q17$, as domestic residents have longer systemic exposure to state legislation.
  • Execution: Replacing $Q17$ with $Q16$ in pivot table columns yields nearly identical column percentages ($93\%$ vs. $7\%$ and $65\%$ vs. $35\%$) due to high variable overlap between international status and overseas birth.
• Visualization Clean-up Procedure:
  1. Copy pure pivot table values and paste them as raw values into an isolated cell block.
  2. Delete non-response/blank rows and grand total lines to isolate clean comparative data.
  3. Highlight isolated cells -> Navigate to Insert -> Recommended Charts -> Select Clustered Column or Stacked Column.
  4. Access Chart Elements menu -> Toggle Data Labels to display exact integer percentages above or inside column segments.

Questions & Discussion

• Concluding Exchange:
  • Mateo: "Thanks, Mateo. Bye bye."