Did it change before and after — one pre/post set

Raw data 사전사후_프로그램_만족도.csv · 120 rows

Synthetic practice data — not a real survey and not any organization’s data. A seeded generator in the repository builds the 120-row sample: the two groups start from the same place on the pre-test, and only the size of the change differs (the planted coefficients are written in that generator). Every table and APA sentence is a measured result produced by running the repository’s own engines.

The study, step by step — what was actually produced

1. ① Who answered — count participation, gender and age band

   Count first: how many took part and how many did not, and how gender and age band split. Skip this and every later test reports without saying how many people it ran on.

   A frequency analysis was conducted on group, gender and age_band (N = 120). For group (2 categories), "Not participated, Participated" each appeared equally often (60, 50.0% of valid cases); valid n = 120, missing = 0. For gender (2 categories), "Male, Female" each appeared equally often (60, 50.0% of valid cases); valid n = 120, missing = 0. For age_band (4 categories), "20s, 30s" each appeared equally often (32, 26.7% of valid cases); valid n = 120, missing = 0. No missing data were present.

2. ② The shape of the score — before and after, side by side

   Mean, standard deviation, median, range, skewness and kurtosis for both time points at once. A distribution piled up on one side changes which test belongs in the next step.

   Descriptive statistics were computed for satisfaction_pre and satisfaction_post (N = 120). For satisfaction_pre, M = 3.22, SD = 0.58, Mdn = 3.25, range = 1.80–4.80. Skewness = 0.04 and kurtosis = -0.34, within normal range. For satisfaction_post, M = 3.49, SD = 0.71, Mdn = 3.50, range = 1.80–5.00. Skewness = -0.18 and kurtosis = -0.24, within normal range. No missing data were present.

3. ③ Paired-samples t-test — did the same people change

   The same people were measured twice, so this is a paired test, not a comparison of two groups. The mean difference with its confidence interval and an effect size (Cohen’s d) come with it. The engine qualifies when there is one set of paired numeric measures such as pre and post (N≥10), and that is read off the column names — which is why this sample names its columns satisfaction_pre and satisfaction_post.

   Paired-samples t-tests examined change in satisfaction_pre–satisfaction_post (N = 120). There was a significant change from satisfaction_pre (M = 3.22, SD = 0.58) to satisfaction_post (M = 3.49, SD = 0.71), t(119) = -7.53, p < .001, d = -0.69, 95% CI [-0.34, -0.20].

4. ④ The nonparametric alternative — when normality is doubtful

   If the distributional assumption is uncomfortable, work with ranks instead (Wilcoxon signed-rank). It answers the same question by another route, and neither is called better — both passed the eligibility check.

   A Wilcoxon signed-rank test indicated a significant change from satisfaction_pre (Mdn = 3.25) to satisfaction_post (Mdn = 3.50), W = 855.500, Z = -6.254, p < .001, r = .57, N = 120.

5. ⑤ Control the pre-test, then compare groups — ANCOVA

   “Did the participants gain more?” cannot be answered by comparing post-test scores alone, because the starting points may differ. Putting the pre-test in as a covariate compares the groups after leveling those starting points. In the registry this engine sits under group difference, not change over time — but with pre/post data this is where that question gets answered.

   An ANCOVA examined the effect of group on satisfaction_post, controlling for satisfaction_pre (N = 120). The covariate satisfaction_pre was significant, F(1, 117) = 362.71, p < .001. After adjustment, the effect of group was significant, F(1, 117) = 52.92, p < .001, η² = .10. Adjusted means were Group Not participated 3.27, Group Participated 3.71.

Key results

Representative quotes

Every quote is a sentence that exists in the raw data (excerpt-existence check passed).

Paper-ready tables