Notes on Cross-Sectional, Longitudinal, and Mixed Methods Research
Time dimension in research design
Research design often hinges on whether data are collected at a single point or across multiple time points; this time dimension drives how we can infer relationships and causality.
Key terms to remember:
Cross sectional
Longitudinal
The instructor emphasizes revisiting these terms and understanding their differences, strengths, and weaknesses for exam readiness.
Cross-sectional studies
Definition: Cross sectional data are collected at one point in time; essentially a snapshot of what is happening at that moment.
Example given:
End of semester survey of all Troy students on their level of alcohol consumption (one point in time).
Strengths:
Easier and quicker data collection
Less resource-intensive than longitudinal designs
Weaknesses:
Difficult to establish temporality or causality because you cannot determine which variable came first
Associations exist, but directionality and causation are unclear
Practical note from the lecture:
Cross sectional data are common and useful but require careful interpretation about causality; one may state that variables are associated and logically related, but not claim causation without longitudinal evidence or strong theoretical justification.
Longitudinal studies
Definition: Data are collected at multiple points in time; allows examination of changes and temporal ordering.
Qualitative or quantitative data can be collected over time.
Types of longitudinal studies discussed:
Trend studies
Definition: Longitudinal in time, but data at each time point come from different samples of people.
Example data sources:
American Community Survey (A CS): every few years, a new sample of around representative people
General Social Survey (GSS): typically a new sample in each wave, often around people
Purpose: to observe changes in population attitudes or behaviors over time, not changes within the same individuals
Strengths:
Captures population-level shifts over time
Weaknesses:
Cannot track how a given individual’s attitudes change; cannot attribute causality to individuals’ experiences
Descriptive trend examples from lecture:
Attitudes about whether it is okay for teenagers to have sex outside marriage (ages 14–16 in the example) and whether attitudes have become more permissive or less permissive over time. The trend shows a gradual change in the proportion who think “not wrong at all” about teen sex since 1985, with potential deviations around 2014.
Attitudes toward cheating (sex with someone other than a spouse) over time show a complex pattern: likely more permissive since the 1970s/80s, with fluctuations later.
Male vs. female differences in attitudes toward cheating: men more likely to think cheating isn’t always wrong; women more likely to think it is always wrong. These differences are discussed in the context of a trend study.
Additional discussion points:
Education level and attitudes: higher education sometimes associated with more permissive attitudes toward sex with others outside marriage; potential explanations include religious influences and other confounding variables.
The role of spurious variables and the caution needed when interpreting relationships in trend data.
Overall takeaway: Trend studies are useful for observing broad changes over time across different samples, but they don’t track the same individuals.
Cohort studies
Definition: Longitudinal, but data points involve following the same subpopulation over time as it ages, using different samples at each wave that represent that subpopulation.
Example concept:
Following people who were in high school (aged 14–18) at the time of September 11, and then tracking similar age groups a decade later (e.g., ages 24–28, then 34–38) to study generational differences.
Distinguishing feature:
Still uses different groups of people at each time point, but the samples are drawn to represent the same subpopulation cohort across time.
Illustration from the lecture:
A diagram (Figure 4.5) shows liberalism levels by cohort, indicating a generational shift (Gen Z appears more liberal on average in the plotted data).
Practical implications:
Cohort studies help examine how particular subpopulations evolve over time while maintaining a focus on a shared origin (the cohort).
Limitations:
While following a cohort, you still rely on samples at each wave rather than the exact same individuals, which can introduce cohort-specific confounders.
Panel studies
Definition: Longitudinal, but the same individuals are followed and measured repeatedly over time.
Core idea: To study changes within individuals as they experience life events or transitions.
Example imagined in lecture:
Following the same couples pre- and post weight-loss surgery to examine shifts in relationship satisfaction and dynamics over time.
Strengths:
Provides rich, within-person evidence of change and causal sequences when time ordering is clear.
Key challenge:
Panel attrition (panel mortality): participants dropping out over time, either voluntarily or involuntarily (e.g., death, loss to follow-up).
Practical considerations:
Addressing and planning for attrition is essential; dropout can bias results if the attrition is systematic.
Mixed methods approaches
Definition: Research designs that incorporate both quantitative and qualitative components.
Rationale:
Quantitative data offer breadth and generalizability, but may miss depth and context.
Qualitative data provide depth and insight, but may lack generalizability.
Example from lecture:
Start with a quantitative survey of around students to measure stress, GPA, and family background; then select a subsample of about to for qualitative interviews to gain deeper understanding of the factors behind observed patterns.
How they complement each other:
Use quantitative results to identify broad patterns and then use qualitative inquiry to explain mechanisms, meanings, and context behind those patterns.
Practical caveat:
Mixed methods can be very demanding and require careful integration of findings from both components.
The research design process: a cyclical reality
A common academic depiction is a linear sequence: formulate a research question, write a literature review, collect data, analyze, report.
The instructor emphasizes that real research is often cyclical and iterative:
You may start with a question, read literature, refine the question, gather more literature, adjust methods, and then collect data.
Even after data collection, findings can prompt further refinement of the question or methods.
Takeaway: Treat research design as an evolving process rather than a strict linear path.
Practical considerations for longitudinal data
Time dimension matters: if your research question involves change over time or temporal ordering, a longitudinal design is often appropriate.
Data management challenges specific to longitudinal designs:
Maintaining contact with participants over long periods
Linking data across waves (consistent identifiers, data integrity)
Managing large, complex datasets with repeated measures
A candid caveat from the lecture:
Working with longitudinal data can be technically challenging—described humorously as a 'pain in the ass' for the plumbing involved in data collection and management.
Key takeaways for exam readiness
Cross-sectional vs longitudinal: understand definitions, strengths, weaknesses, and what each design can (and cannot) answer.
Know the four longitudinal types discussed: trend, cohort, panel, and the qualitative/quantitative flexibility within longitudinal designs.
Recognize issues unique to longitudinal work, especially panel attrition, and strategies to mitigate them.
Understand mixed methods as a way to balance breadth and depth, with concrete example of combining a survey with qualitative interviews.
Be prepared to discuss why the research design is often cyclical rather than strictly linear, and how this affects planning and interpretation.
Real-world relevance: longitudinal designs enable causal inference through temporal ordering, while cross-sectional designs are valuable for snapshot descriptions and exploring associations.
Quick questions to test understanding (reflective prompts)
If your research question includes a time component, which design is most appropriate and why?
How would you handle potential panel attrition in a 10-year panel study?
What are the strengths and limitations of using a trend study to examine attitudes about sex outside marriage?
How can mixed methods help when a quantitative survey shows a strong association but qualitative data reveal context that explains why the association exists?