Make an X-ray result reproducible

XRISM observations are identified by records that connect the target, acquisition interval and instrument setup. The catalogue can include approved targets as well as completed observations. A planned pointing is not yet a usable exposure, and a public image release is not the same thing as a complete analysis-ready dataset. Check what is available before promising a measurement.

The mission’s archive FAQ explains that commissioning observations can use settings chosen for instrument checkout. Their presence in an observation log does not guarantee suitability for every science question. An analyst should record the observation identifier and relevant operating conditions, not rely on a target name alone.

Start with calibrated events and select intervals, detector elements and event grades appropriate to the analysis. Compare a source model through the instrument response with the observed counts, accounting for background. Preserve the software and calibration versions so another reader can understand why your inferred line position or brightness has a particular value.

In an original simplified example, a selected band contains 120 counts and a background estimate, already scaled to the same area and exposure, is 20 counts. The net estimate is 100 source counts. The subtraction does not make the result exact: both the measured total and the estimated background have uncertainty. If the background was measured over twice the area, using its raw count directly would be a scaling error.

Keep the analysis chain reproducible

A reproducible report states the question, observation identity, selection rules, calibration, background treatment, model and uncertainties. It distinguishes a measured feature from the interpretation placed on it. “This line profile is consistent with this velocity model under these assumptions” is more useful than declaring an entire physical mechanism proven by one plot.

Imagine two researchers obtaining slightly different velocities from the same target. First compare observation intervals and regions, energy calibration, line identities and models. They might be measuring different gas, applying different corrections or encountering a genuine limitation. Keeping those choices with the result makes disagreement investigable instead of reducing it to a contest between headlines.