Fitted curves

One row per fitted analyte. Each fit is a censored maximum-likelihood fit of a shedding curve to one biomarker in one specimen from one study, so the same study appears once per analyte and once per model.

The table opens on analytes reported as concentrations. Analytes reported as cycle thresholds are here too, behind the first filter: a Ct peak is measured in cycles below a reference rather than as a log₁₀ concentration, so the two are never shown together unless you ask for it.

• Analyte is what makes a row unique, and it is the identifier the package uses. A study can measure the same biomarker in the same specimen several ways — a different vaccine dose, gene target, assay, or patient subgroup — and those differ only in this column.
• Day 0 is the event every time in the fit is measured from. Studies anchor to symptom onset, enrollment, vaccination, inoculation and others, so curves are only directly comparable when this matches.
• Peak is the population mean peak, always read against the unit in the next column. For a concentration fit it is a log₁₀ concentration. For a cycle-threshold fit it is cycles below the reference, so a peak of 14 against a reference of 40 is a peak Ct of 26. Higher means more shedding on both scales, but the two are different quantities and never comparable to each other.
• σ is the residual scale — larger means the individual measurements scatter further from the fitted curve.
• Limit is the censoring limit, below which a measurement counts as a non-detect rather than a number. It is on the same scale as Peak, so on a cycle-threshold fit it is also cycles below the reference — a limit of 3 against a reference of 40 means the assay stopped counting at Ct 37.
• AIC is not how to choose between the three models. They are fitted to different observation sets — gamma discards every reading at t ≤ 0 — and AIC only compares fits to the same data, so check Meas. before reading anything into it. Across different analytes it means nothing at all.

Rows are sorted by study and analyte to begin with; any column header re-sorts, and clicking it again reverses. Sorting keeps the current filters applied.

To see a fit plotted against the underlying measurements, open the study with Explore from the datasets page.