Exam Prep Notes: Research Variables and Study Design

Understanding Independent and Dependent Variables in Research

Structure of Research Studies and Variables

• Initial Question Regarding Study Design (Article 1 Example):

  • A common question arises as to why a study wouldn't just use one independent variable (IV) with multiple levels (e.g., "Group" with four levels: Early activity with behavior, Early activity without behavior, Usual activity with behavior, Usual activity without behavior).

  • Limitation of Single IV Approach: This approach would only allow for comparisons between the specific four groups/cells (the "inside question"). It would not allow for broader comparisons across primary intervention types (e.g., comparing all early activity groups vs. all usual activity groups) or across behavior modification conditions (e.g., all behavior plus vs. all behavior minus groups).

• Authors' Chosen Design (Article 1 Example - Correct Approach):

  • The authors did not set up the study with one IV with four levels.

  • They treated Treatment (Usual or Early activity) as one independent variable, with two levels: Early Activity (EA) and Usual Activity (UA).

  • They treated Behavior (Behavior modification) as another independent variable, with two levels: Plus (with behavior modification) and Minus (without behavior modification).

  • This creates a $2 \times 2$ factorial design, forming a table (or matrix) with four cells, each representing a distinct group (e.g., EA + Plus, EA + Minus, UA + Plus, UA + Minus).

• Advantages of the Chosen Design:

  • Allows for broader comparisons by collapsing across levels of one IV to compare levels of another.

  • If there were $25$ people in each of the four cells (total $100$ participants):

    • The researcher can compare all Early Activity participants ($50$ people) to all Usual Activity participants ($50$ people) to answer the broad question about the effectiveness of early intervention. This was the authors' primary research question.

    • Similarly, the researcher can compare all Behavior Plus participants ($50$ people) to all Behavior Minus participants ($50$ people).

  • Additionally, it still allows for specific comparisons between the individual cells (e.g., EA+Plus vs. UA+Plus), which are the same comparisons a single IV approach would yield.

  • This design provides a more comprehensive understanding by examining main effects (due to each IV separately) and interaction effects (how the IVs combine).

Dependent Variables and Their Types (Article 1 Example)

• Balance Scores:

  • Measurement: Number of errors made.

  • Type: Ratio variable.

  • Explanation: A true zero is possible (e.g., $0$ errors). The relationship between scores is meaningful; $4$ errors is twice as many as $2$ errors. This allows for statements like "made twice as many."

• Compliance:

  • Measurement: Compliant or Not Compliant.

  • Type: Nominal variable.

  • Explanation: Categories have no inherent order. It's not possible to be "twice as compliant." Data is typically represented by counts within each category (e.g., $15$ compliant, $12$ non-compliant).

• Pediatric Quality of Life Inventory:

  • Measurement: Graded items on a $4$-point scale.

  • Type: Ordinal variable.

  • Explanation: The scale provides an ordered ranking (e.g., $4$ is better than $2$), but the intervals between points are not necessarily equal, and there's no true zero. A score of $4$ doesn't mean twice the quality of life as a score of $2$. It merely indicates a directional difference or ranking.

Understanding the Concept of a True Zero

• Ratio Variables: Possess a true, meaningful zero point, indicating the complete absence of the measured attribute. Examples include:

  • The number of errors made (can have $0$ errors).

  • Physical measurements like height, weight, thickness, or strength (can have $0$ strength or $0$ thickness).

  • The number of calories eaten today (can be $0$).

• Interval Variables: Have ordered categories with equal intervals between them, but no true zero point. A zero on an interval scale does not mean the absence of the attribute.

  • Example: $0$ degrees Fahrenheit does not mean there is no temperature; it's simply a point on the scale.

  • Rarity in Research: Interval scales are less commonly encountered in typical journal articles compared to nominal, ordinal, or ratio scales. If struggling to classify a variable, it is often not an interval variable unless it's a very specific, well-known psychological scale without a true zero.

Identifying Variables in Research Articles

• Keywords: Look for phrases like "a comparison between…" which often directly indicate the independent variables (e.g., soccer players vs. football players; ACL reconstruction vs. non-reconstruction; specific treatments vs. controls).

• Title: The title itself often reveals what the study is comparing or investigating.

• Methods and Results Sections: Detail what was measured and how, providing clues for dependent variables and their types. Data tables (e.g., comparison tables or outcome measures) are excellent resources.

Analysis of Study 2: Blood Flow Restriction (BFR) Training

• Independent Variables (IVs):

  • IV 1: Type of Training Intervention

    • Levels:

      1. CE (Conventional Exercise)

      2. CE + BFR (Conventional Exercise plus Blood Flow Restriction)

  • IV 2: Time

    • Levels:

      1. $4$ weeks (Measurement at $4$ weeks)

      2. $12$ weeks (Measurement at $12$ weeks)

• Study Design Structure (Conceptual): Similar to the first article, this creates a factorial design where measurements for each training group are taken at two different time points. This allows for comparison of training types, comparison across time, and examination of how training types interact with time.

• Dependent Variables (DVs):

  • Quadriceps Strength:

    • Type: Ratio variable.

    • Explanation: A true zero point exists (no strength). Meaningful ratios and differences can be calculated (e.g., percentage improvement, one person being twice as strong).

  • Quadriceps Thickness / CSA (Cross-Sectional Area):

    • Type: Ratio variable.

    • Explanation: A true zero point exists (no thickness/area). A $2$-inch thick steak is indeed twice as thick as a $1$-inch thick steak.

• Study Findings:

  • BFR Efficacy: Blood Flow Restriction (BFR) did not make a statistically significant difference in strength or size when added to conventional exercise. (Note: "Not significant" means any observed difference could be due to random chance, not a true effect).

  • Training Duration Efficacy: Eight (presumably the difference between $4$ and $12$ weeks, or total $12$ weeks given the context) weeks of training did make a significant difference in strength or size. There was a demonstrable change in dependent variables between the $4$-week and $12$-week measurements.

  • Group Response: Both groups (CE and CE+BFR) showed similar responses over time; the difference was primarily due to training duration, not the BFR addition itself.

Research Study Designs and Evidence Hierarchy

• Systematic Review vs. Meta-Analysis:

  • Systematic Review: Involves searching databases, synthesizing findings from multiple studies, and drawing conclusions based on the conclusions of those studies. It's like counting how many studies support one viewpoint versus another.

  • Meta-Analysis: Takes a systematic review a step further by actually extracting and combining the raw data from individual studies into larger pools. This allows for new statistical analyses on a much larger dataset.

  • Key Difference: Systematic reviews combine conclusions; meta-analyses combine data.

  • Power: Meta-analyses are more powerful because combining data from multiple studies (increasing the "sample size") increases the likelihood of detecting a true effect if one exists.

• Compilations: Both systematic reviews and meta-analyses are collectively referred to as Compilations.

• Hierarchy of Evidence (Pyramid Structure): Compilations sit at the top of the evidence hierarchy, indicating they are the strongest form of evidence.

  1. Compilations (Systematic Reviews, Meta-Analyses)

  2. RCTs (Randomized Controlled Trials)

  3. Cohorts (Cohort Studies)

  4. Case Control / Case Studies (Case-Control Studies, Case Series, Case Reports)