PS1007 #6 Descriptive and Inferential Statistics & SPSS Practice

Measurement, Distributions, and Core Statistical Principles

  • Levels of Measurement:

    • Nominal: Categorical data with no inherent order, quantitative distance, or numerical scale (e.g., experimental condition assignment).

    • Ordinal: Ranked categories where the order indicates relative magnitude, but intervals between adjacent ranks are neither equal nor known.

    • Interval: Quantitative scale with standardized, equal units of distance between consecutive points, but lacking an absolute, true zero point (e.g., degrees Celsius).

    • Ratio: Quantitative scale possessing both equal intervals between points and an absolute, non-arbitrary true zero point. For example, reaction time measured in milliseconds (ms\text{ms}) is a ratio scale because 0 ms0\,\text{ms} represents the complete absence of time, and intervals are strictly equal.

  • Distributions and Skewness:

    • Positively Skewed Distribution (Right-Skewed): A distribution where most scores cluster at the lower end of the scale, and the tail extends toward higher positive values on the right.

    • Negatively Skewed Distribution (Left-Skewed): A distribution where most scores cluster at the higher end of the scale, and the tail extends toward lower negative values on the left.

    • Leptokurtic Distribution: A distribution with high positive excess kurtosis, characterized by a slender, sharp central peak and heavier, fatter tails.

    • Bimodal Distribution: A distribution that exhibits two distinct, separate modes or local peaks.

  • Measures of Central Tendency in Skewed Data:

    • Mean (MM): The arithmetic average, calculated by dividing the sum of all values by the sample size. The mean is sensitive to extreme values and outliers, causing it to be pulled substantially in the direction of the tail in skewed distributions.

    • Median: The middle score when observations are ordered from lowest to highest. The median is resistant to extreme scores. When describing variables with severe positive skew (such as household income where a small percentage of individuals earn disproportionately high amounts), the median serves as the most accurate measure of central tendency for a typical individual.

    • Mode: The most frequently occurring score in a dataset.

  • Descriptive Versus Inferential Roles:

    • Descriptive statistics (e.g., sample mean MM, standard deviation SDSD) summarize, organize, and describe the characteristics of observed sample data.

    • For example, in a comparison where staff on a 4-day working week report burnout scores of M=31.52M = 31.52 (SD=8.20SD = 8.20) and staff on a 5-day working week report burnout scores of M=36.04M = 36.04 (SD=9.11SD = 9.11), descriptive metrics indicate that the 4-day sample scored lower on average and exhibited slightly less variance. However, descriptive statistics cannot establish whether this observed difference exists in the broader population or merely reflects random sampling variability.

    • An inferential statistical test is necessary to determine whether the difference between conditions is larger than would be expected by chance alone.

  • Sampling Distributions and Standard Error:

    • Standard Error of the Mean (SESE): Quantifies the dispersion or standard deviation of sample means around the true population mean across repeated random sampling, calculated as:         SE=SDNSE = \frac{SD}{\sqrt{N}}

    • Impact of Sample Size: Holding all other variables constant, increasing the sample size (NN) decreases the standard error of the mean. A larger sample size allows individual unusual or extreme observations to balance out, causing sample means to fluctuate less and sit closer to the population mean.

Hypothesis Testing, Decision Errors, and Statistical Power

  • The Logic of Null Hypothesis Significance Testing (NHST):

    • Null Hypothesis (H0H_0): Postulates that there is no true effect, association, or difference in the underlying population, and that any observed sample variation arises entirely through chance.

    • Alternative Hypothesis (H1H_1): Postulates that a genuine effect, association, or difference exists in the underlying population.

  • Definition and Interpretation of the pp-Value:

    • The pp-value is the probability of obtaining sample results at least as extreme as those observed, assuming that the null hypothesis is true.

    • A pp-value is not the probability that the null hypothesis is true, nor is it the probability that the alternative hypothesis is true.

    • A pp-value does not measure effect magnitude or practical importance.

  • Decision Rules and Non-Significant Findings:

    • When the obtained probability exceeds the conventional threshold (p>.05p > .05, such as p=.271p = .271), the appropriate statistical conclusion is to fail to reject the null hypothesis.

    • Failing to reject H0H_0 indicates that there is insufficient statistical evidence of a difference. This outcome does not prove that no difference exists, nor does it justify "accepting" the null hypothesis.

    • For example, when evaluating memory recall between students taking practice quizzes (M=14.22M = 14.22, SD=3.10SD = 3.10) and students re-reading notes (M=13.11M = 13.11, SD=3.42SD = 3.42), obtaining t(48)=1.20t(48) = 1.20, p=.238p = .238 indicates no statistically significant difference in recall, precluding any claim that practice quizzes improve memory or that both study strategies are functionally identical.

  • Decision Errors in Hypothesis Testing:

    • Type I Error (False Positive): Occurs when the null hypothesis is rejected when it is actually true in the population. For instance, concluding that a novel therapy reduces anxiety when in reality it produces no effect constitutes a Type I error. The probability of this error is governed by the chosen alpha level (α\alpha).

    • Type II Error (False Negative): Occurs when failing to reject the null hypothesis when it is actually false in the population.

    • True Positive (Statistical Power): Correctly rejecting a false null hypothesis.

    • True Negative: Correctly retaining (failing to reject) a true null hypothesis.

  • Rejection Regions in Directional Versus Non-Directional Tests:

    • In a non-directional (two-tailed) test evaluated at α=.05\alpha = .05, the total 5% rejection region is divided equally between the two tails of the sampling distribution, allocating 2.5%2.5\% (.025.025) into the upper tail and 2.5%2.5\% (.025.025) into the lower tail.

  • Mechanisms of Statistical Power and Noise Reduction:

    • The tt-statistic reflects the ratio of signal (systematic mean difference) to noise (unexplained variation or standard error of the difference):         t=DifferenceNoiset = \frac{\text{Difference}}{\text{Noise}}

    • Holding the mean difference and sample size identical across studies (e.g., a 6-point reduction on an exam stress scale between mindfulness and control conditions), a study with low within-group variability (smaller SDSD, tight clustering) will generate a larger tt-statistic and a correspondingly smaller pp-value than a study with substantial within-group variability (larger SDSD, broad dispersion).

    • Power Advantages of Within-Subjects Designs: Repeated measures (within-subjects) designs are generally more statistically powerful than between-subjects designs because each participant serves as their own baseline. Individual differences (such as baseline cognitive ability or enduring personality traits) remain constant across conditions and cancel out mathematically, reducing error variance (noise) in the denominator.

Parametric Assumptions: Normality and Variance Homogeneity

  • Normality Assumption Criteria:

    • Parametric inferential procedures like the tt-test assume that continuous sample data are drawn from an approximately normally distributed population.

    • Paired-Samples tt-Test Requirement: The assumption of normality does not require individual condition scores to be normally distributed in isolation. Instead, the assumption requires that the difference scores (Condition A−Condition B\text{Condition A} - \text{Condition B}) for each participant follow an approximately normal distribution.

    • Independent-Samples tt-Test Requirement: The assumption of normality must be assessed and met separately within each independent group.

  • Formal Statistical Tests of Normality:

    • Shapiro-Wilk Test: Optimal for assessing normality in smaller and moderate sample sizes (n<50n < 50 observations per group). For an independent-samples design with N=52N = 52 split equally into two groups of n=26n = 26, the Shapiro-Wilk test is the appropriate test because each group contains fewer than 50 observations.

    • Kolmogorov-Smirnov Test (with Lilliefors Correction): Typically applied when group sizes exceed 50 observations.

    • Interpreting Normality Tests: The null hypothesis (H0H_0) posits that the sampled population is normally distributed. A statistically significant result (p≤.05p \le .05, such as p=.003p = .003) indicates that the distribution departs significantly from normality, violating the assumption. A non-significant result (p>.05p > .05, such as p=.999p = .999) indicates that the observed distribution does not differ significantly from a normal distribution, supporting the tenability of the normality assumption.

  • Homogeneity of Variance and Levene's Test:

    • The independent-samples tt-test assumes that the two comparison populations exhibit equal variances (homoscedasticity).

    • Levene's Test for Equality of Variances: Evaluates the null hypothesis that group variances are equal.

    • Output Selection Rule in SPSS:

      • If Levene's test is non-significant (p>.05p > .05), group variances do not differ significantly; interpret the "Equal variances assumed" row.

      • If Levene's test is statistically significant (p≤.05p \le .05, such as p=.021p = .021), group variances differ significantly; interpret the "Equal variances not assumed" row (Welch's tt-test correction).

Design Architectures: Independent Versus Paired-Samples t-Tests

  • Independent-Samples tt-Test:

    • Design Configuration: Between-subjects design involving two distinct, mutually exclusive groups of participants assigned to separate levels of an independent variable (e.g., 40 students randomly assigned to either a sleep-deprivation condition with n=20n = 20 or a normal-sleep condition with n=20n = 20, measuring word recall).

    • Degrees of Freedom (dfdf): Calculated as the total number of participants minus the number of estimated group means:         df=Ntotal−2=n1+n2−2df = N_{\text{total}} - 2 = n_1 + n_2 - 2

    • For a study with 2525 participants per condition (n1=25n_1 = 25, n2=25n_2 = 25, total N=50N = 50), the degrees of freedom are:         df=50−2=48df = 50 - 2 = 48

  • Paired-Samples (Repeated-Measures) tt-Test:

    • Design Configuration: Within-subjects design where the same cohort of participants is tested under two distinct conditions, or where pairs of participants are matched on relevant extraneous variables.

    • Degrees of Freedom (dfdf): Calculated based on the total number of paired difference scores:         df=Npairs−1df = N_{\text{pairs}} - 1

    • For an experiment tracking 2424 participants across two sequential testing intervals, the degrees of freedom are:         df=24−1=23df = 24 - 1 = 23

  • Interpreting Standard Statistical Output Notation:

    • In APA reporting format, test statistics are presented as t(df)=valuet(df) = \text{value}, p=valuep = \text{value}.

    • In the reporting string t(58)=−3.72t(58) = -3.72, p<.001p < .001:

      • The number enclosed within the parentheses (5858) represents the degrees of freedom.

      • The value −3.72-3.72 represents the obtained tt-statistic.

      • The value p<.001p < .001 denotes the associated probability under the null hypothesis.

    • In a report where a sleep-hygiene group slept longer (M=7.40 hoursM = 7.40\,\text{hours}, SD=0.80SD = 0.80) than a control group (M=6.75 hoursM = 6.75\,\text{hours}, SD=0.90SD = 0.90), yielding t(38)=2.41t(38) = 2.41, p=.021p = .021, the result demonstrates a statistically significant group difference at α=.05\alpha = .05, warranting the rejection of the null hypothesis.

Mock Quiz: Questions, Options, and Detailed Rationales

  • Question 1: Level of Measurement for Reaction Time

    • Question: Reaction time measured in milliseconds is an example of which level of measurement?

    • Options: 1) Nominal, 2) Ordinal, 3) Interval, 4) Ratio.

    • Correct Answer: 4) Ratio.

    • Rationale: Reaction time possesses equal intervals between units and a meaningful absolute zero (0 ms0\,\text{ms} denotes the complete absence of time).

  • Question 2: Central Tendency for Positively Skewed Income

    • Question: Household income is strongly positively skewed, with a few people earning very large amounts. Which measure of central tendency best describes the typical income?

    • Options: 1) Mean, 2) Median, 3) Mode, 4) Range.

    • Correct Answer: 2) Median.

    • Rationale: The median is resistant to extreme values in the tail, whereas the mean is distorted by extreme positive outliers.

  • Question 3: Identifying Positive Skew from a Histogram

    • Question: A histogram shows most scores bunched at the low end, with a long tail stretching out to the right. This distribution is:

    • Options: 1) Positively skewed, 2) Negatively skewed, 3) Leptokurtic, 4) Bimodal.

    • Correct Answer: 1) Positively skewed.

    • Rationale: The direction of skew is defined by the direction of the elongated tail; a tail stretching toward positive values indicates positive skew.

  • Question 4: Role of Descriptive Versus Inferential Statistics

    • Question: Staff on a 4-day week reported burnout scores of M=31.52M = 31.52, SD=8.20SD = 8.20. Staff on a 5-day week reported M=36.04M = 36.04, SD=9.11SD = 9.11. Based only on these descriptive statistics, which statement is correct?

    • Options: 1) Staff on a 5-day week are definitely more burnt out in the population; 2) The 4-day group's scores are more spread out than the 5-day group's; 3) The 4-day group scored lower on average, but we need an inferential test to know whether this difference is bigger than we would expect by chance; 4) The two groups have the same average burnout.

    • Correct Answer: 3) The 4-day group scored lower on average, but we need an inferential test to know whether this difference is bigger than we would expect by chance.

    • Rationale: Descriptive statistics quantify only sample values. Inferential statistics are required to evaluate whether sample differences exceed expected chance variation.

  • Question 5: Identifying Components of Inferential Output

    • Question: A results section reports: t(58)=−3.72t(58) = -3.72, p<.001p < .001. What does the number 58 show?

    • Options: 1) The t-statistic, 2) The degrees of freedom, 3) The p-value, 4) The number of participants in each group.

    • Correct Answer: 2) The degrees of freedom.

    • Rationale: The integer located in parentheses immediately following the test symbol tt represents the degrees of freedom (dfdf).

  • Question 6: Definition of a pp-Value

    • Question: What does a p-value tell you?

    • Options: 1) The probability that the null hypothesis is true; 2) The probability that the alternative hypothesis is true; 3) How big the difference between the groups is; 4) The probability of getting results at least as extreme as ours, if the null hypothesis were true.

    • Correct Answer: 4) The probability of getting results at least as extreme as ours, if the null hypothesis were true.

    • Rationale: A pp-value assumes that H0H_0 is true; it does not indicate the posterior probability of H0H_0 or the size of an effect.

  • Question 7: Interpreting a Non-Significant Result

    • Question: A researcher obtains p=.271p = .271. Which statement is correct?

    • Options: 1) We have proven there is no difference between the groups; 2) We accept the null hypothesis; 3) We fail to reject the null hypothesis: there is no evidence of a difference, but this does not show there is no difference; 4) We reject the null hypothesis.

    • Correct Answer: 3) We fail to reject the null hypothesis: there is no evidence of a difference, but this does not show there is no difference.

    • Rationale: When p>.05p > .05, the decision is to fail to reject H0H_0. In scientific testing, the null hypothesis is never proven or accepted.

  • Question 8: Identifying Type I Decision Errors

    • Question: A researcher concludes that a new therapy reduces anxiety (rejects H0H_0), but in reality the therapy has no effect. This is a:

    • Options: 1) Type I error (false positive), 2) Type II error (false negative), 3) True positive, 4) True negative.

    • Correct Answer: 1) Type I error (false positive).

    • Rationale: Rejecting a true null hypothesis constitutes a Type I error (false positive).

  • Question 9: Partitioning Alpha in Two-Tailed Tests

    • Question: In a two-tailed test with α=.05\alpha = .05, how is the 5% rejection region split?

    • Options: 1) All 5% sits in the upper tail; 2) 5% in each tail (10% in total); 3) 2.5% in each tail; 4) 0.5% in each tail.

    • Correct Answer: 3) 2.5% in each tail.

    • Rationale: A non-directional test divides the nominal alpha level equally across both tails of the distribution (.052=.025\frac{.05}{2} = .025 per tail).

  • Question 10: Interpreting a Significant Shapiro-Wilk Test

    • Question: The Shapiro-Wilk test for one group gives p=.003p = .003. What does this suggest?

    • Options: 1) The data are normally distributed; 2) The data are not normally distributed; 3) The two groups have equal variances; 4) There is a significant difference between the groups.

    • Correct Answer: 2) The data are not normally distributed.

    • Rationale: A significance level of p≤.05p \le .05 rejects the null hypothesis of normality, indicating significant departure from a normal distribution.

  • Question 11: Sample Size and Standard Error of the Mean

    • Question: Everything else being equal, if you increase your sample size, the standard error of the mean will:

    • Options: 1) Get larger, 2) Stay the same, 3) Become equal to the standard deviation, 4) Get smaller.

    • Correct Answer: 4) Get smaller.

    • Rationale: With increased sample size, atypical scores balance out, reducing variation among sample means and causing them to cluster more tightly around the true mean.

  • Question 12: Drawing Plain-English Conclusions from Non-Significant Tests

    • Question: A study reports: "Students who tested themselves with practice quizzes recalled slightly more words (M=14.22M = 14.22, SD=3.10SD = 3.10) than students who re-read their notes (M=13.11M = 13.11, SD=3.42SD = 3.42), t(48)=1.20t(48) = 1.20, p=.238p = .238." Which plain-English conclusion is correct?

    • Options: 1) There was no significant difference in recall between the two groups, so we cannot conclude that practice quizzes improve memory; 2) Practice quizzes significantly improve memory compared with re-reading; 3) Re-reading notes is better for memory than practice quizzes; 4) The study proves that both methods are equally effective.

    • Correct Answer: 1) There was no significant difference in recall between the two groups, so we cannot conclude that practice quizzes improve memory.

    • Rationale: Because p>.05p > .05, there is no evidence of an effect. This outcome cannot be interpreted as proof that the two study methods are equivalent.

  • Question 13: Normality Assumption in Paired-Samples Designs

    • Question: For a paired-samples t-test, what needs to be approximately normally distributed?

    • Options: 1) The scores in each condition, checked separately; 2) The difference scores (condition A minus condition B); 3) The independent variable; 4) The total of both conditions added together.

    • Correct Answer: 2) The difference scores (condition A minus condition B).

    • Rationale: The paired-samples tt-test evaluates each participant's difference score; therefore, normality must hold for these computed differences.

  • Question 14: Mechanism of Power in Within-Subjects Designs

    • Question: Why are within-subjects (repeated measures) designs usually more powerful than between-subjects designs?

    • Options: 1) They always use larger samples; 2) They do not need a null hypothesis; 3) Each person acts as their own baseline, so individual differences cancel out and there is less noise; 4) They do not require any assumptions to be checked.

    • Correct Answer: 3) Each person acts as their own baseline, so individual differences cancel out and there is less noise.

    • Rationale: Comparing individuals against their own baseline removes extraneous between-person variability, reducing the error term (noise).

  • Question 15: Selecting Tests for Two Independent Groups

    • Question: Researchers randomly assign 40 students to either a sleep-deprivation group (n=20n = 20) or a normal-sleep group (n=20n = 20), then count how many words each student recalls. The data are normally distributed in both groups. Which test is appropriate?

    • Options: 1) Paired-samples t-test, 2) Wilcoxon signed-rank test, 3) Mann-Whitney U test, 4) Independent-samples t-test.

    • Correct Answer: 4) Independent-samples t-test.

    • Rationale: The study involves two independent groups, a continuous dependent variable, and normally distributed data.

  • Question 16: Normality Test Selection by Group Sample Size

    • Question: A study has 52 participants, split equally into two groups. Which normality test should you use?

    • Options: 1) Shapiro-Wilk, because each group has 26 observations (fewer than 50); 2) Kolmogorov-Smirnov, because the total sample is over 50; 3) Levene's test; 4) No test is needed for an independent t-test.

    • Correct Answer: 1) Shapiro-Wilk, because each group has 26 observations (fewer than 50).

    • Rationale: Normality for an independent tt-test must be evaluated separately within each group. With n=26n = 26 per group, the sample size per test is below 50.

  • Question 17: Interpreting Levene's Test Violations

    • Question: When running an independent-samples t-test in SPSS, Levene's test gives p=.021p = .021. Which row of the t-test output should you read?

    • Options: 1) "Equal variances assumed"; 2) Neither: you must switch to a Mann-Whitney U test; 3) Either row: it does not matter; 4) "Equal variances not assumed".

    • Correct Answer: 4) "Equal variances not assumed".

    • Rationale: A Levene's test with p≤.05p \le .05 violates the assumption of equal variances, requiring the use of the corrected "Equal variances not assumed" row.

  • Question 18: Degrees of Freedom Calculation for Independent tt-Tests

    • Question: An independent-samples t-test is run with 25 participants in each group. What are the degrees of freedom?

    • Options: 1) 24, 2) 48, 3) 49, 4) 50.

    • Correct Answer: 2) 48.

    • Rationale: For an independent-samples tt-test, df=Ntotal−2=(25+25)−2=48df = N_{\text{total}} - 2 = (25 + 25) - 2 = 48.

  • Question 19: The Ratio of Signal to Noise in Inferential Testing

    • Question: Two studies both find that a mindfulness group scores 6 points lower on exam stress than a control group, with the same sample sizes. In Study A, scores within each group are tightly clustered (small SD); in Study B they are very spread out (large SD). Which is true?

    • Options: 1) Both studies will give the same t-statistic, because the difference is the same; 2) Study B will have a larger t-statistic and a smaller p-value; 3) Study A will have a larger t-statistic and a smaller p-value; 4) Neither study can be analysed with a t-test.

    • Correct Answer: 3) Study A will have a larger t-statistic and a smaller p-value.

    • Rationale: The tt-ratio is calculated as difference divided by noise. Smaller standard deviations reduce noise, increasing tt and decreasing the associated pp-value.

  • Question 20: Evaluating Statistical Significance in Group Comparisons

    • Question: A study reports: "Participants in the sleep-hygiene group slept longer (M=7.40 hoursM = 7.40\,\text{hours}, SD=0.80SD = 0.80) than the control group (M=6.75 hoursM = 6.75\,\text{hours}, SD=0.90SD = 0.90), t(38)=2.41t(38) = 2.41, p=.021p = .021." Which is the correct interpretation?

    • Options: 1) There was a statistically significant difference in sleep duration between the groups, so we reject the null hypothesis; 2) There was no significant difference, because p is greater than .01; 3) The study proves the workshop makes everyone sleep longer; 4) There is a 2.1% probability that the null hypothesis is true.

    • Correct Answer: 1) There was a statistically significant difference in sleep duration between the groups, so we reject the null hypothesis.

    • Rationale: Because p=.021<.05p = .021 < .05, the effect is statistically significant, warranting the rejection of the null hypothesis.

SPSS Analysis: Physical Activity and Cognitive Performance

  • Experimental Design and Variables (cogperf.sav):

    • Research Objective: Investigate whether physical activity levels maintained across a 2-week intervention period influence cognitive performance.

    • Design Architecture: Within-subjects (repeated measures) design. Every participant completed both experimental conditions, with each condition lasting 2 weeks.

    • Independent Variable (IV): Activity level, operationalized across two levels:

      • Sedentary Condition: Participants took a short walk once daily and did not perform workouts.

      • Active Condition: Participants engaged in structured workouts 3 times per week and took short walks 3 times daily.

    • Dependent Variable (DV): Cognitive performance score, assessed at the conclusion of each 2-week intervention.

    • Counterbalancing: Half of the participants completed the active condition first, while the remaining half completed the sedentary condition first, counterbalancing order effects.

    • Statistical Hypotheses:

      • Alternative Hypothesis (H1H_1): Activity level affects cognitive performance. Cognitive performance will differ between the sedentary and active conditions.

      • Null Hypothesis (H0H_0): Activity level does not affect cognitive performance. There will be no difference in cognitive performance between the sedentary and active conditions.

  • Normality Assumption Testing of Difference Scores:

    • Normality was evaluated for the computed difference scores (difference_scores=Cognitive PerformanceSedentary−Cognitive PerformanceActive\text{difference\_scores} = \text{Cognitive Performance}_{\text{Sedentary}} - \text{Cognitive Performance}_{\text{Active}}) across the sample of N=24N = 24 participants.

    

SPSS Tests of Normality Output
*   **Numerical Normality Tests**:
    *   *Kolmogorov-Smirnov Test*: Statistic = .104.104, df=24df = 24, p=.200∗p = .200^* (with Lilliefors significance correction; indicates a lower bound of the true significance).
    *   *Shapiro-Wilk Test*: Statistic = .992.992, df=24df = 24, p=.999p = .999.
    *   *Interpretation*: The Shapiro-Wilk test is non-significant (p=.999>.05p = .999 > .05). The distribution of difference scores does not depart significantly from a normal distribution, supporting the normality assumption.

    

Histogram of Cognitive Performance Difference Scores
*   **Visual Assessment via Histogram**:
    *   *Observed Parameters*: Sample size N=24N = 24, mean difference M=−2.26M = -2.26, standard deviation SD=3.209SD = 3.209.
    *   *Interpretation*: The histogram shows an approximately bell-shaped, unimodal, and symmetrical distribution centered near −2.00-2.00, with no extreme scores or pronounced skewness. This appearance is consistent with normality.

    

Normal Q-Q Plot of Cognitive Performance Difference Scores
*   **Visual Assessment via Normal Q-Q Plot**:
    *   *Observed Configuration*: Plotted points represent observed difference values against expected normal values along the diagonal reference line.
    *   *Interpretation*: The observed values fall close to the straight reference line across the full range of data, with only a slight deviation at the lowest observation. This is consistent with a normal distribution.

    

Boxplot of Cognitive Performance Difference Scores
*   **Visual Assessment via Boxplot**:
    *   *Observed Configuration*: Displays the median line, interquartile box (25th25\text{th} to 75th75\text{th} percentiles), and whiskers.
    *   *Interpretation*: The horizontal median line sits near the center of the box, the upper and lower whiskers are roughly symmetrical, and no outliers or extreme scores appear beyond the whiskers. This pattern is consistent with normality.
  • Execution and Interpretation of the Inferential Test:

    • Test Selection: Because the design is within-subjects and the difference scores satisfy the assumption of normality, a paired-samples tt-test was conducted.

    • Sample Descriptives:

      • Active Condition: M=104.00M = 104.00, SD=5.95SD = 5.95

      • Sedentary Condition: M=101.73M = 101.73, SD=5.24SD = 5.24

    • Inferential Test Statistics: t(23)=−3.45t(23) = -3.45, p=.002p = .002

    • Statistical Decision: Because p=.002<.05p = .002 < .05, the null hypothesis is rejected.

  • Standard APA Results Paragraph Structure:

    • A complete APA results write-up follows five core reporting steps:

      1. Justify: State the assessment of assumptions (e.g., normality checks) justifying the chosen test.

      2. Name: Explicitly identify the specific statistical test conducted.

      3. Descriptives: Report the means (MM) and standard deviations (SDSD) for each condition.

      4. Inferential: Report the test statistic, degrees of freedom, and exact pp-value (formatted as t(df)=valuet(df) = \text{value}, p=valuep = \text{value}).

      5. Meaning: Provide a plain-English substantive interpretation of the findings.

    • Model APA Paragraph:         Data exploration showed that the difference scores were approximately normally distributed: the histogram was roughly bell-shaped, the dots on the Q-Q plot fell close to the line, the boxplot was symmetrical with no outliers, and the Shapiro-Wilk test was not significant (p=.999p = .999). Therefore, a paired samples t-test was conducted to compare cognitive performance between the sedentary and active conditions. Participants showed higher cognitive performance after the active condition (M=104.00M = 104.00, SD=5.95SD = 5.95) than after the sedentary condition (M=101.73M = 101.73, SD=5.24SD = 5.24). This difference was statistically significant, t(23)=−3.45t(23) = -3.45, p=.002p = .002. This suggests that being more physically active over a 2-week period may improve cognitive performance.