Source dataset
The copied data dictionary identifies the source dataset as kaistrain.dta.
KAIS 2012 reproducible analytics example
This public example follows the same order a user should expect in Reproducible Analytics Studio: prepare the dataset, check it with EDA, build Table 1, build Table 2, build Table 3, and then generate the AI-assisted manuscript and export package.
KAIS 2012
13,614 records
26 variables
Tables 1-3
AI-assisted manuscript
Expected app workflow
Read this example in the same order you would run the app.
The useful part is not just the final manuscript. The workflow keeps each source output visible, so a reviewer can trace the report back to the dataset, checks, tables, model output, figures, and export files.
1
Data preparedness
Confirm the source dataset, records, variables, labels, value distributions, and missingness before using the results.
2
EDA
Use the data dictionary, observed distributions, and descriptive checks to understand the data before formal tables.
3
Table 1
Describe participant characteristics by sex so the study population is clear before interpreting outcomes.
4
Table 2
Estimate HIV prevalence by sex and subgroup, with weighted percentages and confidence intervals.
5
Table 3
Review simple and multiple logistic regression estimates for predictors of HIV positivity.
6
AI report and export
Generate the manuscript-style draft only after the checked source outputs are available for review.
Data Preparedness
The example starts with the source dataset and variable map.
Before any table is treated as meaningful, the user needs to know what dataset was loaded, how many records are present, and what each variable represents.
Dataset shape
The example includes 13,614 records and 26 variables in the source data dictionary.
Variable review
The output keeps field names, labels, formats, missingness, and observed values available before the tables are interpreted.
First file to open
Start with the data dictionary when reviewing this example, then move to the analysis tables.
Open data dictionary →EDA
The data is checked before the main analysis tables.
In this public package, the EDA evidence is mainly visible through the data dictionary and the descriptive distribution outputs. The goal is to catch coding, missingness, and distribution issues before the formal tables are read as final.
Observed values
The data dictionary shows observed value distributions so a reviewer can see how variables are coded.
Missingness review
Missingness is kept close to the variable descriptions instead of being separated from the analysis trail.
Descriptive distributions
The manuscript package also carries a descriptive distribution figure from the saved frequency output.
Public-package limit
The example does not expose a full EDA dashboard. It shows the copied outputs that are available for review now.
Table 1
The first formal table describes the study participants.
Table 1 helps a reviewer understand the study population before moving to prevalence or predictor results.
Table title
Participant characteristics by Sex, with N=13,614.
Survey settings
The output records aiweight, strata2, and qclust for the survey-style table.
Visible details
The table keeps weighted estimates, confidence intervals, p-values, footnotes, and metadata visible.
Review next
Use Table 1 to check participant structure before interpreting Table 2 prevalence results.
Open Table 1 →Table 2
The second table answers the HIV prevalence question.
Table 2 estimates HIV prevalence among study participants by sex and subgroup, with weighted percentages and confidence intervals.
5.68%Overall weighted HIV prevalence, 95% CI 4.98-6.37
6.92%Weighted prevalence among women
4.39%Weighted prevalence among men
15.21%Weighted prevalence in Nyanza among listed regions
Outcome definition
The prevalence output uses hiv = 1 as the outcome-positive definition.
Analyzable denominator
The table reports 648 HIV-positive records among 11,534 analyzable participants for the selected output.
Design metadata
The output records abweight, strata2, and qclust, plus footnotes about approximation and weighting.
Open the table
Review the full table before relying on the short prevalence summary.
Open Table 2 →
Table 3
The third table keeps predictor evidence visible.
Table 3 shows simple and multiple logistic regression estimates together, so a reviewer can see both unadjusted and adjusted results before reading the manuscript draft.
11,534Model-output N in the copied predictor table
4.69Nyanza adjusted OR, 95% CI 2.92-7.51
1.44Women adjusted OR, 95% CI 1.18-1.76
Visible caveatApproximate weighted-GLM and robust-covariance notes remain in the output.
| Predictor signal | Simple estimate | Adjusted estimate | Review note |
|---|---|---|---|
| Nyanza vs Eastern | OR 4.59 | Adjusted OR 4.69 | Strong regional signal in both simple and multiple outputs. |
| Women vs men | OR 1.62 | Adjusted OR 1.44 | Sex remains visible after adjustment in the copied table. |
| Modeling caveat | Weighted output | Robust-covariance approximation | The table notes that it does not reproduce PROC SURVEYLOGISTIC Taylor variance exactly. |
AI Report And Export
The manuscript comes after the checked source outputs.
The AI-assisted manuscript is useful because it pulls the analysis trail into a readable draft. It should still be checked against the data dictionary, Table 1, Table 2, Table 3, and figures before publication or formal use.
The draft covers HIV prevalence and associated factors among adults aged 15-64 in KAIS 2012. It is best read as a reporting workflow example, not as a final journal submission.
What the manuscript package includes
- HTML report for quick review.
- PDF for reading or circulation.
- DOCX for editing and coauthor comments.
- Markdown for source-style review.
- Figures copied from the saved analysis outputs.
Report Outputs
Open the draft and review formats.
PDF and DOCX
Use PDF for review and DOCX when the report needs edits or comments.
Open PDF → Open DOCX →Output Package
The page shows highlights; the source files show the detail.
This page stays readable by showing the path and key values. The linked files are the source evidence for anyone who wants to inspect the details.
Source checks
Data dictionary with labels, formats, observed values, missingness, and source dataset context.
Analysis tables
Table 1 characteristics, Table 2 prevalence, and Table 3 predictors are all available as standalone HTML outputs.
Figures
The manuscript package carries prevalence, descriptive distribution, and logistic-regression figures.
Report package
HTML, PDF, DOCX, and Markdown outputs are kept together for review and reuse.
What is included
A public example, not the full workspace.
The example shows the analysis flow and copied outputs. The live app is where users would upload data, run checks, create tables, generate reports, and manage saved jobs.