Purpose

Public preview: This page shows the analysis output, figures, diagnostics, and reporting guidance without exposing the R code. The paid version includes the complete editable R Markdown source, reusable functions, all analysis code, synthetic data, documentation, and exported results.

This education-specific workflow connects survey cleaning, item review, reliability, factor evidence, scale scoring, weighting, group comparisons, and leadership outcome models. The example follows educators responding to organizational climate, trust, and efficacy items. It is designed as a decision workflow rather than a chain of automatic significance tests.

User settings

Simulate or import

Survey export audit

Structure, duplicates, and codes

Survey structure audit
responses districts duplicate_ids invalid_likert
1105 28 0 0

Missing response patterns

Item missingness
item missing_n percent
climate1 0 0.0
climate2 43 3.9
climate3 0 0.0
climate4 0 0.0
trust1 0 0.0
trust2 0 0.0
trust3 45 4.1
efficacy1 0 0.0
efficacy2 44 4.0
efficacy3 0 0.0

Item distributions

# Coding and scale evidence {.tabset .tabset-pills} ## Reverse scoring and reliability

Internal-consistency estimates
scale scale_alpha standardized_alpha average_interitem
Climate 0.837 0.836 0.560
Trust 0.779 0.779 0.540
Efficacy 0.797 0.797 0.567

Factor-structure review

Three-factor exploratory loading pattern
item MR1 MR2 MR3
climate1 0.84 0.01 0.02
climate2 0.85 0.00 -0.04
climate3 0.71 0.05 0.03
climate4 0.71 -0.05 0.02
trust1 -0.04 0.02 0.87
trust2 0.05 0.01 0.72
trust3 0.03 -0.03 0.70
efficacy1 0.01 0.82 0.01
efficacy2 0.02 0.79 -0.03
efficacy3 -0.03 0.75 0.03

Reliability alone does not validate a scale. Use theory, dimensionality, item wording, response processes, and evidence from a new sample. A weak or cross-loading item should not be deleted solely to maximize alpha.

Scale construction

Nonresponse and weights

Weight distribution and effective sample size
min median max effective_n
0.56 1 1.43 1037.82

Weights require documentation of the sampling frame, selection probabilities, nonresponse adjustment, calibration, trimming, and variance method. The example normalizes supplied synthetic base weights for teaching; it does not estimate nonresponse weights without a frame containing respondents and nonrespondents.

Leadership questions

Adjusted mean innovation

Survey-weighted model of innovation score
term estimate std.error statistic p.value conf.low conf.high
(Intercept) 33.692 1.435 23.473 0.000 30.688 36.696
climate 3.169 0.268 11.836 0.000 2.609 3.730
trust 0.437 0.216 2.027 0.057 -0.014 0.889
efficacy 2.057 0.243 8.460 0.000 1.548 2.566
rolePrincipal 0.341 1.157 0.295 0.771 -2.081 2.763
roleTeacher -0.615 0.865 -0.711 0.486 -2.426 1.195
years_education -0.042 0.048 -0.885 0.387 -0.141 0.057
school_levelHigh -0.283 0.729 -0.387 0.703 -1.809 1.244
school_levelMiddle 0.043 0.613 0.070 0.945 -1.241 1.327

Intent to leave

Weighted proportional-odds model of intent to leave
term odds_ratio low high p.value
climate 0.511 0.456 0.573 0.000
trust 0.695 0.625 0.772 0.000
efficacy 0.918 0.829 1.016 0.098
rolePrincipal 0.968 0.613 1.527 0.888
roleTeacher 0.893 0.611 1.304 0.558
years_education 1.001 0.985 1.018 0.895
school_levelHigh 1.351 1.035 1.764 0.027
school_levelMiddle 1.149 0.885 1.491 0.298

Adjusted predictions

# Multiple testing and small groups

Define a small number of primary outcomes and contrasts before analysis. Report effect sizes and intervals, not a screen of unadjusted p-values. Collapse identity or role categories only with substantive justification and stakeholder input; otherwise report suppression or uncertainty rather than presenting unstable estimates. Clustering at district or school level and a small number of clusters may require design-specific corrections.

Dynamic reporting example

In the synthetic weighted analysis, a one-point increase in organizational climate was associated with 3.17 points higher innovation, 95% CI [2.61, 3.73], p < .001. In the ordinal model, the corresponding adjusted odds ratio for being in a higher intent-to-leave category was 0.51, 95% CI [0.46, 0.57], p < .001.

Reporting checklist

  • Document the sampling frame, recruitment, response rate, eligibility, duplicates, and skip logic.
  • Preserve labels and distinguish legitimate skips, structural missingness, and item nonresponse.
  • Report item distributions, reliability with limitations, and dimensionality evidence before scoring.
  • State score rules, reverse coding, minimum completed items, and missing-data strategy.
  • Describe weights, clustering, strata, finite-population corrections, and effective sample size.
  • Match the model to the outcome and report adjusted means or probabilities with coefficients.
  • Address small groups, multiplicity, nonresponse bias, common-method bias, and causal limits.

When this workflow is not enough

Use confirmatory factor analysis and measurement invariance for strong latent-variable claims; multilevel models for school and district variance; multiple imputation for defensible incomplete-data analyses; longitudinal models for change; and qualitative or mixed-methods designs when leadership meaning and context are central. Cross-sectional self-report associations do not establish leadership effects.

Export results