hillSlope1 (the spread link's slope) and inflectionPoint1 (its Richards exponent) are fixed at 1, not fitted (fireSenseUtils::fixedLogisticPars; see estimateSpreadParams()): neither is identifiable together with the covariate coefficients, so they are no longer part of DEoptim's parameter space, and paramsBest (from bestParamSets()) does not include them either. fireSense_spreadPredict splits a ledger row's parameters from its covariates BY NAME, and the spread link (fireSenseUtils::logistic3p()/logistic3pUpper()) reads the result BY POSITION – maxAsymptote, hillSlope1, inflectionPoint1, and (with the upper-tail link) upperTail1. This restores those positions for new fits, so a new ledger row predicts with both at 1 exactly like an old row whose hillSlope1 and inflectionPoint1 happened to be fitted at that value, and an old row keeps predicting with its own fitted values.

addFixedParsToLedger(paramsBest)

Arguments

paramsBest

a data.table, one row per parameter set, as bestParamSets()$params returns it: columns named names(P(sim)$lower), so without hillSlope1 and inflectionPoint1, maxAsymptote first.

Value

paramsBest with hillSlope1 and inflectionPoint1 columns of 1s inserted right after maxAsymptote.