One resident survey, end to end: questionnaire to report

Raw data citizen_perception_survey_2026.csv · 12 rows

Synthetic demonstration data — not a real survey and not any municipality’s data (the city is written as a placeholder). The 1,000-row sample comes from a seeded generator in the repository, and every table and sentence was produced by running the repository’s own engines. Those values were then checked cell by cell against an independently computed answer sheet: 22 cells, 0 mismatches.

The study, step by step — what was actually produced

1. ① Intake — who answered

   Count first: how many fall in each gender, district and investment priority, and where cells are missing. Skip this and every later test reports without saying how many people it ran on.

   A frequency analysis was conducted on gender, district and _priority_investment (N = 1000). For gender (2 categories), the most frequent category was "Male" (506, 50.6% of valid cases); valid n = 1000, missing = 0. For district (4 categories), the most frequent category was "Central" (273, 27.3% of valid cases); valid n = 1000, missing = 0. For _priority_investment (6 categories), the most frequent category was "Transport" (210, 21.0% of valid cases); valid n = 1000, missing = 0. No missing data were present.

2. ② Preprocessing — flip the reverse items, build the scales

   Two reverse-coded items are flipped back, then each scale becomes the mean of its items. The flip uses the same one-line formula the preprocessing screen uses. Then the shape of overall satisfaction and the four scale means is checked.

   Descriptive statistics were computed for _overall_satisfaction and 4 other variables (N = 1000). For _overall_satisfaction, M = 3.24, SD = 1.27, Mdn = 3.00, range = 1.00–5.00. Skewness = -0.17 and kurtosis = -1.01, within normal range. For admin_service_mean, M = 3.37, SD = 0.87, Mdn = 3.40, range = 1.00–5.00. Skewness = -0.17 and kurtosis = -0.62, within normal range. For living_environment_mean, M = 3.22, SD = 0.89, Mdn = 3.20, range = 1.00–5.00. Skewness = -0.19 and kurtosis = -0.64, within normal range. For civic_participation_mean, M = 2.87, SD = 0.91, Mdn = 3.00, range = 1.00–5.00. Skewness = 0.08 and kurtosis = -0.56, within normal range. For trust_mean, M = 3.02, SD = 0.97, Mdn = 3.00, range = 1.00–5.00. Skewness = -0.00 and kurtosis = -0.71, within normal range. 16 missing value(s) were identified.

3. ③ Reliability — do these items hang together

   Cronbach alpha checks whether the five administrative-service items really measure one thing. Item-total correlations and alpha-if-deleted come back with it.

   Internal consistency of Administrative service was Cronbach’s α = .81 (5 items, N = 910). Corrected item-total correlations ranged from .52 to .67. Standardized α = .81; mean inter-item correlation = .46.

4. ④ Factor analysis — do they split into four as designed

   All eighteen items go in at once to see how many groups they form. KMO and Bartlett say first whether the data support factoring at all; eigenvalues and cumulative explained variance follow.

   An exploratory factor analysis with varimax rotation was conducted. The Kaiser-Meyer-Olkin measure verified the sampling adequacy, KMO = .90; Bartlett’s test of sphericity was significant, χ²(153, N = 603) = 3685.10, p < .001. 4 factor(s) were retained, explaining 58.8% of the total variance. Factor 1 was primarily defined by q12, q11, q14. Factor 2 was primarily defined by q7, q8, q6. Factor 3 was primarily defined by q2, q1, q4. The analysis was based on N = 603. Factor loadings ranged from .61 to .80. Factor 1 explained 13.5% of the variance, Factor 2 explained 15.3% of the variance, Factor 3 explained 15.5% of the variance, Factor 4 explained 14.5% of the variance. Communalities ranged from .39 to .73.

5. ⑤ Group comparison — does living environment differ by district

   With four groups you run one ANOVA, not three t-tests. When the effect is significant, the post-hoc test (Tukey HSD) names which district pair separated.

   One-way ANOVAs examined living_environment_mean across levels of district (N = 1000). For living_environment_mean, there was a significant effect, F(3, 996) = 8.52, p < .001, η² = .03. Levene's test indicated equal variances (p = .890). Post hoc Tukey HSD comparisons showed East differed from New town (p = .021), East differed from Central (p = .024), West differed from New town (p < .001), West differed from Central (p < .001). Group statistics: East (M = 3.12, SD = 0.88), West (M = 3.04, SD = 0.87), New town (M = 3.36, SD = 0.91), Central (M = 3.35, SD = 0.87).

6. ⑥ Regression — what raises overall satisfaction

   The four scale means predict overall satisfaction. Standardized beta gives the priority order across domains, and VIF answers whether the predictors are too much alike.

   A multiple linear regression was conducted to predict _overall_satisfaction from admin_service_mean, living_environment_mean, civic_participation_mean, trust_mean. The overall model was significant (R² = .59, adjusted R² = .59, F(4, 979) = 355.56, p < .001). The predictors accounted for 59.2% of the variance in _overall_satisfaction. admin_service_mean significantly predicted _overall_satisfaction (B = 0.54, SE = 0.04, β = .37, t(979) = 14.46, p < .001). living_environment_mean significantly predicted _overall_satisfaction (B = 0.37, SE = 0.03, β = .26, t(979) = 11.15, p < .001). civic_participation_mean significantly predicted _overall_satisfaction (B = 0.25, SE = 0.03, β = .18, t(979) = 7.49, p < .001). trust_mean significantly predicted _overall_satisfaction (B = 0.29, SE = 0.03, β = .22, t(979) = 8.72, p < .001). Multicollinearity was assessed using the variance inflation factor; the maximum VIF was 1.56 (admin_service_mean), indicating no serious multicollinearity.

Key results

Representative quotes

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

Paper-ready tables