Harmonisation Data
From the numbers as printed to the numbers the forest needs — with the calculation shown
One trial gives median [IQR], another mean ± SE, a third hours instead of minutes. Converting by hand is slow, and the reviewer will ask how you did it.
You give
The Extraction workbook (Reconciled sheet) or your own sheet with one row per study × outcome × group (n, mean, SD/SE/CI, median, IQR/range, events, %, unit).
You get
The harmonised workbook with method, worked formula and unit conversion on every row, the Units sheet (target unit and factor per outcome) and the data handling notes in English, ready for the supplement, with references.
How it works
# Harmonisation $ scopetools run harmonize Deterministic — no model. Median → mean by Luo 2018 and SD by Wan 2014 (scenarios C1/C2/C3); SE → SD = SE × √n; CI → SD with t when n < 60 (Cochrane 6.5.2); % → rounded events. Units: time, mass, volume, pressure, temperature, analyte concentrations (molar mass) and 0–10 ↔ 0–100 scales, with the factor cited. Ambiguous cases (no n, no known factor) are left blank with the reason — never guessed. Reported values stay next to the converted ones.
Deterministic; it takes the extraction sheet (raw, as printed), converts each row with the Cochrane-recommended methods — Luo 2018 for the mean and Wan 2014 for the SD from a median, SE × √n, CI with the t distribution when n < 60, % to rounded events — brings each outcome to one unit with a cited factor, and leaves the ambiguous cells blank with the reason. The worked formula sits next to every value, and the data handling notes come out in English, ready for the supplement.
What it costs
3 credits per spreadsheet (US$0.15), whatever the number of rows.
Charged only when it delivers — a failed run costs nothing.
Credits are shared across every MetaScope tool — buy once, spend anywhere.
Limits
It does not decide comparability for you: units without a known factor and medians without n stay blank for you to resolve on the page.
Frequently asked questions
- Which method for the median?
- Luo et al. 2018 for the mean and Wan et al. 2014 for the SD, as the Cochrane Handbook (6.5.2.5) recommends; Wan for the mean is an option.
- Does it guess when data are missing?
- No; a median without n or a unit without a known factor stays blank with the reason.
- What do I paste in the paper?
- The handling notes: one paragraph per conversion, with the studies and the references.
Harmonised? Check the sheet with Data check before pooling — 3 credits.
