Skip to contents

Convert a multi-observer capture history table to a two-observer CR table

Source: R/callDensitySupport.R
mchToCR.Rd

Extracts a two-observer (table1, table2) capture-recapture capture history table from a multi-observer source table by selecting only rows where at least one of the two named observers detected the event, and renaming the relevant columns to the _table1/_table2 suffixes expected by the rest of the callDensity package. The verdict column from the source is used as the ground-truth detection for table1.

Usage

mchToCR(d, table1suffix, table2suffix)

Arguments

d

A multi-observer capture history data.frame containing per-event columns t0, tEnd, fLow, fHigh, SNR, signalRMSdB, noiseRMSdB, noiseDev, verdict, plus per-observer detection columns named detect_<observerSuffix>.

table1suffix

Character. Suffix of the observer columns to remap to the _table1 role (treated as ground truth via verdict).

table2suffix

Character. Suffix of the observer columns to remap to the _table2 role (the detector being evaluated).

Value

A capture history data.frame in the two-observer (table1, table2) format expected by cde, with time/season columns added by capHistTimeSeason.

Deprecated

No current analysis needs this reduction. cde()'s only use of capHistTab is falseDiscoveryRate(), which already accepts gtColName/testColName directly against native detect_observerN columns (once cde() forwards them – see its own argument list). fitDetFun(modelType = 'vglm') accepts any number of native-named observer columns directly via yColNames, with no reduction to two required at all – VGAM::posbernoulli.t is not limited to two occasions. This function's body is unchanged and will keep working, kept only so the published Common Ground and Beyond Counting Calls analysis scripts keep reproducing exactly. New work should skip the 2-observer reduction entirely and use chtToSNRinfo with a observers vector of whatever length is needed.