Skip to contents

Check that recorded fits can be reproduced

Source: R/fit_provenance.R
check_reproducibility.Rd

Re-run each recorded fit with refit_from_provenance and compare the new rows with the stored ones.

Usage

check_reproducibility(
  se,
  assay = NULL,
  fit_source = NULL,
  tolerance = sqrt(.Machine$double.eps),
  allow_source = FALSE
)

Arguments

se

SummarizedExperiment

assay

string or NULL; check only fits written to this assay.

fit_source

string or NULL; check only this fit source.

tolerance

numeric; relative tolerance for numeric columns, as in all.equal.

allow_source

logical; evaluate fit functions stored as source code.

Value

data.table with one row per checked fit: assay, fit_source, reproducible (NA when the fit could not be re-run), n_rows and detail.

Examples

mae <- get_synthetic_data("finalMAE_small.qs2")
se <- mae[["single-agent"]]
se <- apply_fit(
  se, fit_drug_response_metrics, "single-agent",
  output_assay = "custom_hill", fit_source = "hill"
)
#> Warning: NaNs produced
#> Warning: NaNs produced
#> Warning: NaNs produced
#> Warning: NaNs produced
#> Warning: NaNs produced
#> Warning: NaNs produced
#> Warning: NaNs produced
#> Warning: NaNs produced
#> Warning: NaNs produced
#> Warning: NaNs produced
#> Warning: NaNs produced
#> Warning: NaNs produced
#> Warning: NaNs produced
#> Warning: NaNs produced
#> Warning: NaNs produced
check_reproducibility(se)
#>          assay fit_source reproducible n_rows        detail
#>         <char>     <char>       <lgcl>  <int>        <char>
#> 1: custom_hill       hill         TRUE    200 results match