Exhaustive Notes on 2x2 ANOVA Interpretation and SPSS Data Reporting

Descriptive Statistics vs. Inferential Statistics in Data Analysis

• Conceptual Overview:     * Descriptive Statistics: Summarize the data as it exists. These are found on the first page of the SPSS data output and include metrics like the mean ($M$) and standard deviation ($SD$).     * Inferential Statistics: Used to make predictions or test hypotheses; they determine if the observed differences between groups are statistically significant.

• Factorial Design Groups ($2 \times 2$ Matrix):     * The study analyzes two independent variables (Job Status and Type of Attire).     * Main Effect 1 (Job Status): Assistant versus Executive.     * Main Effect 2 (Attire Comparison): Conservative versus Sexy.     * Interaction Groups: The data is broken down into four distinct crossed conditions:         * Assistant-Conservative ($AC$)         * Assistant-Sexy         * Executive-Conservative ($EC$)         * Executive-Sexy

• Standard Deviation and Error:     * The standard deviations are relatively consistent across all groups, which indicates a low amount of unplanned error in the data.

Sample Size ($n$) and Degrees of Freedom ($df$)

• Total Sample Size: The total number of human participants reported in the methods section is $n = 77$.

• Determining Degrees of Freedom ($df$):     * In a $2 \times 2$ ANOVA, there is one degree of freedom per effect.     * Since there are three effects total ($2$ main effects and $1$ interaction), three degrees of freedom are subtracted from the total sample size for the calculation of significance.

• Adjusted N for Reporting:     * Though the total $n = 77$, the adjusted number used in reporting (the error term for the $F$-statistic) is $n = 74$ because of the three subtracted degrees of freedom ($77 - 3 = 74$).     * Reporting Exception: A former colleague named Arvivostaxian occasionally reported the intercept, resulting in four degrees of freedom, though this is considered non-standard practice in the department.

• Statistical Power: Removing degrees of freedom does not mean removing participants; it is a mathematical adjustment that helps the statistics account for possible error and provides more power during the ANOVA calculation in SPSS.

APA Formatting and Result Reporting Standards

• Italics and Spacing:     * Statistical symbols such as $F$, $M$, $SD$, and $p$ must be italicized.     * There should be no space between the symbol and the parentheses surrounding the degrees of freedom (e.g., $F(1, 74)$).

• Decimal Precision and Rounding:     * $F$-scores: Report with two digits after the decimal (e.g., $57.84$).     * $p$-values: Report with three digits after the decimal (e.g., $p = 0.041$) to show as much significance as possible.     * Rounding Rule: If the third or fourth digit is a $5$, round the preceding digit up.

• Anatomy of a Significance Statement:     * Example: "We found a significant main effect of job status, $F(1, 74) = 57.84$, p < 0.001."     * The first number in the parentheses ($1$) indicates the degree of freedom for the main effect, and the second ($74$) is the corrected total sample size.

• Explaining Directions of Effects:     * After reporting the test statistic, descriptive statistics ($M$ and $SD$) must be provided to show the direction of the finding.     * Example from Data: Perception of competence for Executives ($M = 50.84$, $SD = 4.60$) was higher compared to Assistants ($M = 43.91$, $SD = 4.56$).

• Hypothesis Confirmation: The final sentence of a results block should state whether the hypothesis for that specific effect was supported.

Inferencing and Measurement in Psychology

• Direct vs. Indirect Measures:     * Indirect (Inferential): Psychology and Sociology often use indirect measures (like Likert scales) because we cannot see the mind directly; we must infer behavior from data.     * Direct (Neuroscience): Uses invasive methods like placing electrodes in specific structures (e.g., Broca's area) to measure neural firing. This requires fewer statistics because it is a direct measurement.

• Cognitive Experiment Example: Identifying real words (e.g., Tree, House) versus non-words.     * Non-word examples: "GLIC" and "PIPL." These are used because they follow vowel/consonant patterns that look like real words, forcing the brain to process them.     * Dependent Variables ($DV$):         * $DV_1$: Reaction times measured in milliseconds ($ms$).         * $DV_2$: Accuracy measured as percent correct.

Interpreting Interactions via Graphics

• Interaction Definition: An interaction occurs when the effect of one independent variable depends on the levels of another independent variable.

• Visual Representation:     * Parallel Lines: Indicate that no interaction is present; the effects are independent.     * Crossing/Converging Lines: Indicate an interaction. Perfect interactions form an "X" shape, but any approach toward crossing is significant.

• Current Study Results: The interaction in this dataset was "subtle" with a significance of $p = 0.04$. While close to the alpha threshold of $0.05$, it is still considered statistically significant.

• Bar Graphs: Can be manipulated for clarity. Starting the Y-axis at a higher number (e.g., $35$ instead of $0$) can make differences between bars look more dramatic and highlight the interaction effect.

Survey Construction: Composite Scores and Reverse Scoring

• Composite Scores: The total score for an individual participant across all items of a measurement tool.     * Process: If a participant takes a $10$-item Likert scale and answers "$5$" on every item, their composite score is $50$. The average of all participants' composite scores becomes the group Mean ($M$).

• Reverse Wording: Items phrased in the opposite direction of the construct being measured.     * Purpose: To ensure validity and control for Response Set (the tendency for a participant to answer all questions the same way due to fatigue or lack of attention) and lying.

• Reverse Scoring Math: Adjusting the numerical value of a reverse-worded item so it aligns with the rest of the scale.     * Formula: Take the Likert scale maximum plus one, then subtract the participant's answer ($X$).     * For a $4$-point scale: $5 - X$. (e.g., a score of $4$ becomes a $1$, and a $1$ becomes a $4$).     * Study Specifics: Item #7 in the survey packet was the reverse-scored item.

Homogeneity of Variance and Levene's Test

• Definition: Checking if the variance (spread) of scores is similar across all experimental conditions.

• Heterogeneity vs. Homogeneity:     * Researchers want homogeneity within groups (consistency among participants in a single condition).     * Researchers want heterogeneity between groups (clear differences between the experimental conditions).

• Levene's Test: A specific statistical test used to check for homogeneity of variance. It ensures there is no "weird crazy error" in one specific group compared to others.

Participant Demographics and Recruitment

• Total Participants: $n = 77$.

• Recruitment: Participants were recruited from psychology classes; consequently, females outnumber males in the sample.

• Age Profile:     * Mean Age ($M$): $23$ years.     * Age Range: $18$ to $55$ years.

• Discussion Section Use: Detailed breakdowns of student majors/careers can be used in the discussion section to suggest directions for future research or limitations of the study.

Questions & Discussion

• Q: Does the adjusted n of 74 go in the results or methods?     * A: The total sample of $77$ goes in the methods section under "Participants." Use the adjusted number ($74$) when reporting the $F$-statistic degrees of freedom in the results section.

• Q: When rounding significant digits, do we round up for a 5?     * A: Yes, rounding up is the traditional approach. It is generally preferred because higher numbers are often desired in statistical reporting.

• Q: Why do we use reverse wording?     * A: It is primarily to control for a "response set," where a tired or bored participant might just mark "Agree" for every item without reading the content. It ensures the item is measuring their actual perception.

• Q: What is the purpose of Levene's Test?     * A: It checks for homogeneity of variance to make sure the experimental groups are comparable and that variance is not skewing the results.