Sociological Research Methods: Quantitative Analysis, Correlation, and Causality

Epistemological Foundations and Types of Knowledge

• Empirical Knowledge versus Unprovable Convictions:

  • Empirical inquiry relies on systematic data collection to understand patterns and implement interventions that improve the social world.

  • Non-empirical domains encompass powerful, unprovable realities such as faith, tradition, and spiritual conviction, which hold equal personal and social significance.

  • Sociological and scientific inquiry does not require the elimination of faith or tradition; empirical analysis and personal belief coexist in human behavior.

• Interpersonal Evidence and Trust:

  • Belief in unobservable interpersonal constructs (such as knowing a family member, like a mother, loves their child) is evaluated through longitudinal behavioral evidence.

  • This knowledge relies on accumulated evidence: consistent historical support, predictable responses in times of need, communicative affirmation, and mutual trust built across time.

• Methodological Skepticism and Data Operationalization:

  • Persistent public and intellectual skepticism toward science and data collection necessitates transparent research methodologies.

  • Data represents measurable properties of reality, formally structured through variables.

  • Methodological approaches fall into two primary traditions:

    • Quantitative Research Methods: Numerical measurement, mathematical modeling, and statistical analysis of variables.

    • Qualitative Research Methods: Interpretive, descriptive, contextual analysis of human experiences and social meanings.

  • Operationalization: Complex social and psychological concepts (e.g., mental health, life satisfaction) require explicit operational definitions to define exactly how variables are measured, ensuring independent researchers can replicate studies precisely.

Levels of Statistical Analysis: Univariate, Bivariate, and Multivariate

• Univariate Analysis:

  • Involves the analysis of a single variable ($1$ variable) in isolation.

  • Serves as the most fundamental form of data analysis.

  • Excludes any consideration of cause-and-effect relationships or associations between independent ($x$) and dependent ($y$) variables.

  • Functions primarily to describe and summarize the baseline properties of a dataset.

  • Common descriptive metrics include central tendency measures such as the mean or median.

  • Calculation of the arithmetic mean:         $\bar{x} = \frac{\sum x}{n}$         where all observed values are summed and divided by the total number of observations ($n$).

  • Examples of univariate metrics:

    • Calculating the mean test score of a cohort of students to assess overall academic performance.

    • Examining the median household income of American families, recorded at approximately $\$101,800$ (or about $\$101,000$).

    • Evaluating the average SAT scores of an incoming college student cohort.

• Bivariate Analysis:

  • Examines the relationship, co-variation, or statistical dependence between two ($2$) variables.

  • Designates one variable as the independent predictor ($x$) and the second as the dependent outcome ($y$).

  • Convention places the independent variable ($x$) on the horizontal axis and the dependent variable ($y$) on the vertical axis (following alphabetical precedence where $x$ precedes $y$).

  • Utilizes visual representations such as scatter plots to reveal underlying patterns, trajectories, and associations.

  • Assesses how changes in one variable correspond to changes in another, while maintaining the strict methodological principle that association does not establish causal dependency.

• Multivariate Analysis:

  • Analyzes relationships between three ($3$) or more variables simultaneously.

  • Reflects the complexity of the empirical social world, where single outcomes are rarely governed by isolated bivariate mechanisms (analogous to the multiple variables dictating highway traffic conditions).

  • Structural configurations include:

    • Multiple independent variables predicting a single dependent variable ($x_1, x_2, x_3 \rightarrow y$).

    • A single independent variable influencing multiple distinct dependent variables ($x \rightarrow y_1, y_2$).

  • Multivariate conceptual models:

    • Independent Variable: Level of Education ($x$); Dependent Variables: Life Satisfaction ($y_1$) and Job Satisfaction ($y_2$), which require distinct operational boundaries despite conceptual overlap.

    • Independent Variable: Dietary and nutritional patterns ($x$); Dependent Variables: Serum cholesterol concentration ($y_1$) and blood pressure metrics ($y_2$).

    • Independent Variable: Social media volume/screen exposure ($x$); Dependent Variables: Mental health markers ($y_1$) and academic achievement ($y_2$).

    • Independent Variables: Target audience segmentation ($x_1$), marketing and advertising strategies ($x_2$), physical product features ($x_3$), pricing points ($x_4$), and brand recognition ($x_5$); Dependent Variable: Total commercial product sales ($y$) in profit-driven business enterprises.

Directionality and Strength in Bivariate Correlation

• Foundational Characteristics of Correlation:

  • Correlation quantifies the degree and direction to which two or more variables co-vary.

  • Categorized into three distinct relational states:

    • Positive Correlation: The independent variable ($x$) and dependent variable ($y$) move in parallel; as $x$ increases, $y$ increases.

    • Negative Correlation: The independent variable ($x$) and dependent variable ($y$) exhibit an inverse relationship; as $x$ increases, $y$ decreases.

    • Zero / No Correlation: No systematic linear relationship exists between changes in $x$ and changes in $y$.

• Scatter Plot Interpretation:

  • Data points are plotted along Cartesian coordinates against a trend line (line of best fit).

  • The strength of the correlation is determined by the proximity of the data points to the trend line:

    • Clustering closely along the trend line indicates a strong correlation.

    • Wide dispersion or high distance from the trend line indicates a weak correlation.

• Outliers and Relational Anomalies:

  • Outliers represent atypical data points that diverge sharply from the fitted trend line.

  • In a positive correlation (e.g., human physical dimensions where height positively correlates with weight), outliers represent exceptionally tall individuals with unusually low body mass, or shorter individuals with high body mass.

  • In a negative correlation (e.g., weekly physical exercise hours versus weekly social media duration), the general trend shows that dedicating hours to physical fitness decreases remaining leisure time within a finite $24$-hour daily budget.

  • Outlier analysis in exercise versus media use demonstrates that individuals can exhibit high exercise alongside elevated screen time due to qualitative factors:

    • Consumption of fitness-specific gym media.

    • Working professionally as fitness and athletic influencers.

    • Participating in location-based augmented reality gaming (e.g., Pokémon GO), which historically compelled players to walk extensive distances outdoors while glued to mobile screens.

    • Severe behavioral distractions during the peak of Pokémon GO (which predated TikTok) led to pedestrian traffic fatalities and bridge falls, necessitating programmatic in-game warnings against driving while playing.

• Hypothesizing Study Hours and Academic Outcomes:

  • An increase in study duration is hypothesized to correlate positively with exam grades.

  • Bivariate limitations: Massive study duration does not guarantee high performance; academic outcomes are mediated by prior foundational knowledge, innate cognitive gifts, and continuous exposure.

  • Longitudinal mastery requires extensive deliberate practice ("practice makes perfect"); innate gifts remain undeveloped without continuous task exposure (e.g., the extensive performance hours logged by musical groups such as The Beatles prior to international recognition).

Establishing Causality: The Three Essential Criteria

• Sociological Versus Natural Science Methodologies:

  • Natural sciences frequently isolate mechanisms within controlled laboratory setups (e.g., observing chemical enzymes, proteins, or reagents inside a petri dish).

  • Social sciences (sociology, anthropology) study spontaneous, conscious human actors within uncontrolled aggregate environments (neighborhoods, cities, social groups, institutions).

  • Multiple theoretical frameworks often compete to explain a single human behavior, making strict causal attribution difficult.

• Criterion 1: Empirical Association (Correlation):

  • Two variables must demonstrably co-occur or exhibit statistically significant covariation.

  • Correlation is the mandatory first step toward demonstrating causality, but correlation alone never proves a causal relationship.

  • Domestic Abuse Case Study: While alcohol consumption and domestic partner abuse frequently correlate (batterers often consume excessive alcohol), alcohol abuse cannot be deemed the direct cause of spousal violence. Alcohol acts as an exacerbating or disinhibiting catalyst rather than the primary etiology, contrary to popular moralistic assumptions (where approximately $9$ out of $10$ survey respondents incorrectly infer direct causation).

  • Violent Media Case Study: Sociological consensus demonstrates that consuming violent media correlates with certain aggressive metrics, yet violent media consumption does not directly cause individuals to commit violent shootings or mass crime.

• Criterion 2: Temporal Precedence (Time Order):

  • The presumed causal agent ($x$) must unequivocally precede the presumed effect ($y$) in chronological sequence:         $x \rightarrow y$

  • Bidirectional Causality Dilemma: Many social and behavioral phenomena exhibit reciprocal or circular feedback loops rather than linear time order:

    • Poor sleep duration may induce elevated psychological stress, but elevated psychological stress independently disrupts sleep architecture ($y \rightarrow x$).

    • Engaging in physical exercise can elevate subjective happiness, but elevated happiness provides the motivation to exercise, whereas depressive states cause behavioral withdrawal and inactivity.

• Criterion 3: Non-Spuriousness (Elimination of Confounding Variables):

  • The empirical relationship between $x$ and $y$ must not be produced by an extraneous, unmeasured third variable ($z$), alternatively termed a confounding, lurking, or unseen variable.

  • A spurious correlation is an artificial statistical association generated entirely by the mutual dependence of two variables on an underlying third variable.

  • The Ice Cream and Drowning/Shark Attacks Spurious Relationship: A high, documented positive correlation exists between commercial ice cream sales ($x$) and coastal drowning incidents or shark attacks ($y$).

    • Lurking Variable ($z$): High ambient summer temperatures and seasonal weather changes independently drive increased ice cream consumption and drive greater population density into ocean waters.

  • Family Meal Frequency and Academic Performance Spurious Relationship: A positive correlation exists between children who regularly eat dinner at a dining table with their families ($x$) and high academic test scores ($y$).

    • Lurking Variables ($z$): Household socioeconomic status, parental financial resources (eliminating the need for parents to work secondary shifts), greater available parental time dedication, structured parenting styles, and systemic educational capital.

Logical Fallacies and Cognitive Biases in Causal Reasoning

• The Post Hoc Fallacy (Post hoc ergo propter hoc):

  • Translates from Latin as: "After this, therefore because of this."

  • Logical Formulation: Event $A$ occurs, followed temporally by event $B$; therefore, event $A$ is assumed to have caused event $B$:         $\text{Event } A \rightarrow \text{Event } B \implies A \text{ caused } B$

  • Human cognitive architecture is inherently pattern-seeking, leading individuals to fabricate causal narratives rather than accepting stochastic randomness or chance occurrences.

  • Attribution of Foodborne Illness: Individuals developing gastrointestinal illnesses reflexively attribute symptoms to their most recently consumed distinct meal (e.g., sushi, burritos, pizza).

    • Ignores alternative factors: Prior culinary consumption across the preceding $48$ to $72$ hours, airborne pathogens, pre-existing gastrointestinal viruses, or the development of psychological taste aversions.

    • Contrasts with verified epidemiological food outbreaks where pathogens are isolated systematically (e.g., bacterial contamination of lettuce and tomatoes in Chipotle outbreaks).

  • Athletic and Superstitious Rituals: Superstitious associations where an individual wears specific apparel (e.g., "lucky" running socks or a Real Madrid jersey) followed by consecutive victories (e.g., a $6$-game winning streak), leading the mind to assign causal power to the garment.

  • Vaccines and Autism Spectrum Diagnoses: Standard pediatric vaccine schedules temporally coincide with the developmental window (around $6$ months to toddlerhood) where neurodevelopmental conditions such as autism are clinically detected and diagnosed. The post hoc fallacy leads to the erroneous political and cultural belief that vaccines cause autism, mistaking chronological proximity for biological etiology.

  • Amulets and the Evil Eye: Vehicle owners display traditional evil eye protective talismans and report an absence of motor vehicle accidents, falsely attributing defensive properties to the amulet rather than defensive driving or the psychological placebo effect.

• Confirmation Bias:

  • A cognitive bias whereby individuals selectively register, recall, and emphasize empirical evidence that confirms pre-existing hypotheses or cultural stereotypes, while dismissing, forgetting, or reinterpreting non-conforming data.

  • Freeway Driving Stereotypes: A driver traveling at $80\,\text{mph}$ in the passing lane encounters two motorists traveling at $65\,\text{mph}$. Upon passing, the driver notes that the slow motorists are older individuals, loudly validating the internal bias ("I knew it was an old person"). Instances of younger slow motorists or older fast motorists are filtered out and fail to register.

  • Vehicular Sub-biases: Biases regarding vehicle types (e.g., an underpowered $2$-cylinder vintage Volkswagen van) or assumptions regarding young drivers failing to use vehicular turn signals.

  • Hospital Emergency Ward Folklore: Healthcare staff experiencing chaotic, high-volume emergency shifts look outdoors, observe a full moon, and attribute the surge in patient volume to the lunar cycle (referencing mythological folklore such as werewolves or lunacy), while forgetting chaotic shifts occurring during other lunar phases or calm shifts during full moons.

Questions & Discussion

• Discussion on Unprovable Beliefs:

  • Question/Prompt: What is the "f-word" that describes powerful concepts that cannot be empirically proven?

  • Response: Faith.

• Discussion on Relational Evidence:

  • Question/Prompt: How does an individual prove or know that someone (e.g., a mother) loves them?

  • Response: Knowledge is established through longitudinal behavioral evidence collected over time—demonstrated through continuous emotional support, reliability, trust, and reciprocal communication.

• Discussion on Defining Correlation:

  • Question/Prompt: What does correlation mean in research methods, and what is the standard adage?

  • Response: Correlation describes two events or variables co-occurring simultaneously, leading observers to believe tentatively that one caused the other; the definitive scientific adage states that correlation does not equal causation.

• Discussion on Univariate Examples:

  • Question/Prompt: What serves as an empirical example of a univariate measure in an academic setting?

  • Response: The mean score of student exam grades, calculated by adding all individual test scores together and dividing by the total number of students.

• Discussion on Bivariate Classification:

  • Question/Prompt: Is analyzing the relationship between average age and college level considered univariate or bivariate?

  • Response: It is a bivariate analysis because it measures the relationship between two distinct variables.

• Discussion on Multivariate Interactions:

  • Question/Prompt: Does job satisfaction fall under life satisfaction when evaluating education levels, or can they be differentiated?

  • Response: While work and life satisfaction are closely intertwined, multivariate research can isolate and differentiate the two as distinct dependent outcomes.

• Discussion on Business Variables:

  • Question/Prompt: What independent variables influence the dependent variable of product sales in a commercial enterprise?

  • Response: Target audience identification, marketing execution, product design, pricing strategy, and brand recognition.

• Discussion on Scatter Plot Outliers:

  • Question/Prompt: Why do outliers exist in the negative correlation between exercise hours and social media screen time?

  • Response: Outliers consume fitness-specific content, work as online fitness creators, or participate in mobile gaming like Pokémon GO while exercising outdoors.

• Discussion on Pokémon GO Incidents:

  • Question/Prompt: What occurred historically when players became overly engrossed in Pokémon GO?

  • Response: Players walked off bridges or entered live vehicular traffic, forcing the developers to code safety pop-ups preventing gameplay while driving.

• Discussion on Alcohol and Domestic Spousal Abuse:

  • Question/Prompt: Does scientific data show that alcohol abuse directly causes spousal abuse?

  • Response: No; while positively correlated, alcohol acts as an exacerbating and encouraging factor rather than an absolute direct cause.

• Discussion on Food Poisoning Attribution:

  • Question/Prompt: Why is blaming a recent meal of sushi for illness considered a logical fallacy?

  • Response: The illness could stem from foods consumed earlier, a pre-existing viral infection, or a developing psychological pseudo-aversion rather than the sushi itself.

• Discussion on Talismans:

  • Question/Prompt: Does an evil eye talisman prevent car accidents, or is it a psychological placebo?

  • Response: It operates through superstition and placebo rather than empirical physical protection.

• Discussion on Third Variables in Drowning:

  • Question/Prompt: What explains the positive correlation between ice cream sales and coastal drownings or shark attacks?

  • Response: The lurking variable is seasonal summer weather and air temperature, which drives individuals to both swim in the ocean and purchase ice cream.

• Discussion on Family Dinner Success:

  • Question/Prompt: Why do children who eat family dinners perform better academically?

  • Response: Underlying socioeconomic factors—including parental income, flexible schedules avoiding secondary jobs, active communication, and structured parenting styles—drive both meal frequency and academic performance.

• Discussion on Confirmation Bias in Driving and Hospitals:

  • Question/Prompt: Why do individuals maintain that older drivers are universally slow, or that full moons induce hospital emergency room chaos?

  • Response: Confirmation bias causes observers to register and remember the instances that confirm the stereotype (an older slow driver or a chaotic full moon) while ignoring non-conforming occurrences.

Administrative Notes and Course Logistics

• Paper 1 Evaluation and Feedback:

  • All Paper 1 submissions will be graded and returned prior to the start of the next class meeting.

  • Students must read their written grades and read all instructional comments thoroughly before arriving in class.

• Upcoming Class Agenda:

  • The subsequent class session on Tuesday will expand on the domestic abuse and alcohol case study using formal sociological frameworks.

  • A portion of the session will be dedicated to student sharing regarding their experiences writing Paper 1.

  • Sharing is completely voluntary; past cohorts exhibit an $80\%$ to $90\%$ voluntary sharing rate, with $4$ to $5$ students typically opting out of public sharing.