# Synthetic record-selection flow: methods and figure legend This is a teaching example, not a real observational study or a production model run. All counts were created for this example. They are aggregate counts, not patient-level records. The material contains no observed exposure values, outcomes, identifiers, effect estimates, confidence intervals, or p-values. It must not be submitted as evidence of a real study. ## Question and scope How can a researcher turn a frozen table of sequential exclusion counts into a readable selection-flow diagram and a complete figure legend? The unit is a record. A count of records must not be relabelled as a count of unique people without a verified source definition and deduplication method. This example starts with 1,200 records assessed for eligibility and finishes with 850 included records. ## Input contract `counts.csv` is the complete input table. Its columns are `order`, `stage_id`, `description`, `input_n`, `excluded_n`, `remaining_n`, and `data_status`. Counts are non-negative integers. Every row explicitly carries `synthetic_demonstration`. The first row establishes the starting count; the last row repeats the final included count and is not another exclusion. There is no hidden dataset, patient row, model-generated study finding, or unavailable supplementary file. ## Frozen selection order 1. Start with 1,200 records assessed for eligibility. 2. Exclude 120 records with age below 18 years; 1,080 remain. 3. Among those 1,080 records, exclude 80 with missing exposure; 1,000 remain. 4. Among those 1,000 records, exclude 150 with missing outcome; 850 remain. 5. Label the final box as 850 included records. Do not subtract the final row again. The three exclusion categories are mutually exclusive by this sequential allocation rule, not because their underlying characteristics could never overlap. A record meeting several rules is attributed to its first applicable exclusion only. Moving the order could change each exclusion-category count in a real dataset; this aggregate example cannot recalculate such a change. Preserve the displayed order and do not add unverified independent category totals. ## Deterministic construction and checks The shared English `flow.svg` was laid out deterministically for this example; it was not generated by a model or a ToolRun. Main boxes display the starting, intermediate, and final counts. Side boxes display the exclusions from the current remaining set. Arrows indicate sequence, not causal effects. Before using a flow diagram, check both local and overall arithmetic: `1200 - 120 = 1080`, `1080 - 80 = 1000`, `1000 - 150 = 850`, and `1200 - 120 - 80 - 150 = 850`. Check the unit, frozen order, labels, legend, and matching counts across the input table and editable output. These arithmetic checks do not verify source-data accuracy, actual deduplication, ethical approval, or research validity. ## Figure legend **Figure. Synthetic example of sequential record selection.** Of 1,200 records assessed for eligibility, 120 were excluded for age below 18 years. Among the remaining 1,080 records, 80 were excluded for missing exposure. Among the remaining 1,000 records, 150 were excluded for missing outcome, leaving 850 included records. Exclusion reasons are mutually exclusive under the stated order; each record is counted at its first applicable exclusion. All numbers are synthetic teaching values and do not describe participants in a real study. No effect or outcome analysis is presented. ## Interpretation and limitations The diagram answers how many records were retained under the stated rules. It does not show incidence, treatment benefit, association, causality, selection-bias magnitude, or the quality of missing-data handling. Excluding missing values here is an illustrative selection rule, not a general recommendation to use complete-case analysis. A real study must justify its missing-data strategy and report the relevant study stages, dates, source, unit, and reasons for non-participation or exclusion. Aggregate counts alone cannot test overlaps, identify duplicate records, or audit whether the rules were applied correctly. The page text is localized; these original files and both shared SVG diagrams are in English. Downloading or copying them is not a platform execution or independent scientific review. A reusable project draft should retain the synthetic-data warning, let the user revise the question and supply authorized material, and start only after the user's explicit submission. No completion, statistical result, or submission readiness is promised. ## Method reference STROBE's official checklists provide a reporting reference. Item 13 concerns numbers at study stages, reasons for non-participation, and consideration of a flow diagram. This example uses that general reporting concept; it does not reproduce a STROBE checklist or claim STROBE certification, full checklist compliance, or study-quality endorsement. - STROBE, Checklists: https://www.strobe-statement.org/checklists/ - STROBE, official site: https://www.strobe-statement.org/ - Official checklist page checked on 2026-10-10. Reference access does not validate the synthetic counts. ## Reuse conditions The newly created synthetic `counts.csv`, this methods/legend text, and the original SVG diagrams may be downloaded and adapted for your own teaching or research workflow drafts. Retain an explicit synthetic-data label so that reused numbers cannot be mistaken for observations. This permission covers only these newly created example materials. It does not include STROBE's website, logo, checklists, or any other third-party material; these are not bundled for download.