Steps:

1. Profile each variable: count non-null, distinct, and out-of-range values.
2. Compute missingness and duplicate rates per instrument.
3. Flag variables failing validity rules and document them in a data dictionary.
4. Produce a one-page audit summary table that can be regenerated from code.

Steps:

1. Block records on stable identifiers (email, name, DOB) to reduce comparison space.
2. Apply deterministic exact-match rules first to capture obvious duplicates.
3. Use probabilistic linkage (Fellegi-Sunter) with match weights for fuzzy matches.
4. Clerically review borderline pairs and document the linkage decision rule.

Steps:

1. Visualize the missing-data pattern (e.g., aggr plot, missingness map).
2. Run Little's MCAR test to distinguish MCAR from MAR/MNAR.
3. Decide on a strategy: complete-case, multiple imputation, or sensitivity analysis.
4. Document the assumed missingness mechanism for every key variable.

Steps:

1. Build a codebook documenting each source variable and its scale.
2. Construct a crosswalk mapping each source variable to a target variable.
3. Apply recodes, unit conversions, and category collapses consistently.
4. Version the harmonized dataset and record the pipeline that produced it.

Steps:

1. Store raw data as read-only and never edit it by hand.
2. Write the full pipeline in a version-controlled notebook (Quarto / R Markdown / Jupyter).
3. Set a random seed and declare package versions for stability.
4. Render the final report from a clean environment to confirm reproducibility.

Steps:

1. Compute internal consistency (alpha, omega) for each scale.
2. Estimate test-retest reliability where repeat data exist.
3. Inspect item-total correlations and drop items that weaken the scale.
4. Report reliability with a confidence interval, not a point estimate.

Steps:

1. Specify the measurement model from theory (CFA) or extract factors (EFA).
2. Evaluate fit: CFI/TLI > .95, RMSEA < .06, SRMR < .08.
3. Inspect factor loadings and residual correlations for local dependence.
4. Respecify only with theoretical justification and document each change.

Steps:

1. Choose a model: 1PL/2PL/3PL for binary, GRM for ordinal Likert items.
2. Estimate item difficulty, discrimination, and (if used) guessing.
3. Examine item information and test information curves.
4. Flag items with low discrimination or DIF for review.

Steps:

1. Test configural, then metric, then scalar invariance in sequence.
2. Compare fit across steps; a meaningful drop flags non-invariance.
3. Locate non-invariant items via DIF analysis (e.g., lordif, mirt).
4. Decide on partial invariance or item removal and report the decision.

Steps:

1. Specify the measurement model, then add the structural paths.
2. Estimate with a robust estimator (e.g., MLR) for non-normal data.
3. Evaluate overall and component fit; check path significance.
4. Test alternative models and report a model comparison table.

Steps:

1. Define the event and the time origin; handle right-censoring.
2. Estimate Kaplan-Meier survival curves by group.
3. Fit a Cox model and check the proportional hazards assumption.
4. Report hazard ratios with confidence intervals.

Steps:

1. Specify fixed and random effects (intercepts and slopes).
2. Choose a covariance structure and an appropriate estimator.
3. Model time flexibly (linear, polynomial, or spline).
4. Interpret fixed effects and variance components.

Steps:

1. Fit a latent class growth analysis with increasing numbers of classes.
2. Compare fit (BIC, entropy) and class sizes to choose a solution.
3. Characterize each class by its trajectory shape.
4. Validate classes against external covariates and outcomes.

Steps:

1. Estimate propensity scores and check balance after matching.
2. Apply difference-in-differences with parallel-trends support.
3. Use interrupted time series with a clear intervention point.
4. Report sensitivity analyses for hidden confounding.

Steps:

1. Compare Poisson vs. negative binomial via dispersion tests.
2. Use zero-inflated models when excess zeros are present.
3. Model the dose-response curve (linear, log, or spline).
4. Report incidence rate ratios and predicted counts.

Steps:

1. Familiarize yourself by reading all responses thoroughly.
2. Generate initial codes inductively and/or deductively from the five principles.
3. Group codes into candidate themes and review against the dataset.
4. Define, name, and write up each theme with supporting excerpts.

Steps:

1. Open-code the first transcripts and memo reflexively.
2. Constantly compare new data to emerging codes and categories.
3. Use axial coding to relate categories; selective coding to integrate around a core category.
4. Saturate categories and write a theoretical narrative.

Steps:

1. Define the framework matrix (themes by participants).
2. Index transcripts to the framework categories.
3. Chart data into the matrix and summarize per cell.
4. Interpret across rows and columns, mapping patterns and outliers.

Steps:

1. Draft a codebook with definitions, inclusion/exclusion, and examples.
2. Double-code a sample (e.g., 20%) independently.
3. Calculate Cohen's kappa and reconcile disagreements.
4. Refine the codebook and apply it to the full dataset.

Steps:

1. Clean and tokenize text; remove stopwords and standardize case.
2. Fit a topic model (LDA) and inspect top terms per topic.
3. Run sentiment analysis and examine embeddings for semantic clusters.
4. Triangulate computational findings with manual qualitative themes.

Steps:

1. Define each RE-AIM dimension and its measures for your program.
2. Compute Reach as the proportion of the intended population engaged.
3. Assess Effectiveness and Adoption across settings and staff.
4. Evaluate Implementation fidelity and Maintenance over time.

Steps:

1. Compute rate ratios and rate differences for each group.
2. Estimate a concentration index and plot the concentration curve.
3. Apply Oaxaca-Blinder decomposition to attribute the gap to covariates.
4. Report both absolute and relative measures of inequity.

Steps:

1. Geocode participant and program locations; aggregate to areas.
2. Map choropleths of reach and outcomes.
3. Test for spatial clustering with Moran's I and Getis-Ord Gi*.
4. Use small-area estimation where direct estimates are unstable.

Steps:

1. Disaggregate outcomes by key demographic strata.
2. Examine intersectional subgroups (e.g., race x gender x geography).
3. Test for interaction and report stratum-specific effects.
4. Co-interpret findings with affected community partners.

Steps:

1. Build a community advisory group and agree on evaluation questions.
2. Co-develop measures that are meaningful to the community.
3. Share preliminary findings with the community for interpretation.
4. Co-author recommendations and an action plan for improvement.