An rsnlme_boot object

bootstrap() (and collectJob() on a backgrounded BootNlmeJob) return an rsnlme_boot: a named list of bootstrap result tables with an SCM-style print method and a small set of metadata attributes that downstream tools (notably Certara.Xpose.NLME::get_bootSummaryNlme()) read instead of inspecting the on-disk artefacts directly.

Fields

BootOverall

Per-replicate Scenario / Replicate / ReturnCode / LL.

BootTheta

Aggregated theta summary (Mean, Stderr, CV\ Median, percentile CIs).

BootOmega, BootOmegaCorrelation

Aggregated omega entries split out of BootOmega.csv by the Omega / Correlation sentinel rows.

BootOmegaStderr, BootVarCoVar

Standard errors and full var-covar matrix from the engine.

BootSecondary

Secondary parameters with the same aggregation columns as BootTheta.

BootThetaStacked, BootOmegaStacked, BootSigmaStacked

Per-replicate long stacks (Replicate, <label>, Value). Inputs to transform-aware downstream pipelines.

BootOmegaCI, BootSigmaCI

Per-parameter percentile CIs for omega and sigma (one row per upper-triangle entry). Diagonal marks variance vs covariance entries; print.rsnlme_boot shows omega rows except constrained-zero off-diagonal covariances (matching Original Fit) and diagonal sigma rows only.

BootEtaShrinkage, BootEpsShrinkage

Per-replicate long shrinkages (Replicate, Ranef|Observable, Shrinkage (%)). The CSVs on disk hold fractions (engine convention, 1 - SD) under a Shrinkage header; loadBootstrapResult() multiplies by 100 and renames the column at ingestion so the in-memory representation is on the percent scale. Fractional values only survive in the on-disk artefacts.

fitSummary

Compact prior-fit summary (Parameter, Type, Estimate, SE, %RSE, Shrinkage, Diagonal). Type is one of "the" (theta), "ome" (omega variance or block covariance), "sig" (sigma diagonal), or "sec" (secondary parameter). Estimate follows the prmTable convention used by Certara.Xpose.NLME: raw value for thetas, variance for omega diagonals, covariance for omega off-diagonals, standard deviation for sigmas, raw secondary value (engine output) for sec rows. Diagonal is TRUE except for omega off-diagonal covariance rows (block omega), which downstream variance-scale summaries (Certara.Xpose.NLME::get_summaryNlme()) exclude. Shrinkage is on the percent scale (the .parseEtaShrinkages / .parseEpsShrinkages helpers multiply by 100 at parse time), matching the post-ingestion BootEtaShrinkage / BootEpsShrinkage columns above and Certara.Xpose.NLME::get_summaryNlme()'s output; off-diagonal omega and secondary rows always carry NA since shrinkage is a per-eta concept. The print method labels the column Shrinkage (%) but does not rescale. Present only when the run was launched with initialEstimates = TRUE.

Fields produced by an older NLME8 build that does not yet emit a given aggregate are silently absent (with a single warning at load time).

Attributes

runDir

Path to the directory holding the bootstrap CSVs.

engine

Engine method string derived from params@method.

numSamples

Number of bootstrap replicates requested.

numUsed

Number of replicates contributed to the aggregates, i.e. nrow(BootOverall). Matches NLME8's summarizeBootstrap.R behaviour, which feeds every parseable replicate into the percentile CIs regardless of ReturnCode.

numConverged

Number of replicates whose ReturnCode indicates convergence achieved (c(1L, 2L, 3L) per Certara.Xpose.NLME::get_term). Other codes (e.g. 0, 4, -4, -1) may still contribute parameter estimates to the bootstrap aggregates; see numUsed.

returnCodeBreakdown

Character vector with the per-code occurrence counts (e.g. c("return code 1: 3", "return code -1: 2")), sorted by code. Drives the diagnostic line in print.rsnlme_boot.

confidenceLevel

Confidence level used for percentile CIs (numeric, e.g. 95).

hasFitSummary

TRUE when fitSummary is present.

See also

bootstrap, print.rsnlme_boot, collectJob