Changes:
-
Updating Circe-generated SQL (to avoid errors on DataBricks).
-
Refactoring code so connections do not stay open longer than strictly necessary (to avoid timeout errors on DataBricks).
-
Increasing required version of FeatureExtraction to avoid warning about use of deprecated argument
oracleTempSchema
. -
Added
riskWindowStart
argument tosynthesizeReferenceSetPositiveControls()
for rare occasions when we want to have a specific time-at-risk.
Bugfixes:
-
Fixing nesting cohorts to end at end of observation.
-
Fixing calibration when using exposure controls.
-
Fixing error when computing metrics for negative (or positive) controls only.
-
Using character instead of numeric to store person IDs in R to avoid loss of precision when inserting generated outcomes.
-
Always casting dates to dates before operations to avoid hidden casts to date time.
Changes:
-
Using
checkmate
to check function input. -
Adding
packageCustomBenchmarkResults()
to support custom methods benchmarks. -
Fixing seeds and setting
resetCoefficients = TRUE
to ensure reproducibility of positive control synthesis.
Changes:
-
Updating from
oracleTempSchema
totempEmulationSchema
for newer versions of SqlRender. -
Added unit tests.
Changes:
-
Added 'ohdsiDevelopment' reference set, intended to be used when developing methods. This set is based on 76 negative controls from the LEGEND Hypertension study.
-
Now generating cohort counts when generating reference set cohorts.
Bugfixes:
-
Fixed error during positive control synthesis for platforms using
oracleTempSchema
. -
Adding initial support for 64-bit integer covariate and cohort IDs.Still requires this issue to be resolved.
Changes:
-
Dropping orphan concept and source concept functions, as these are now implemented much better in the CohortDiagnostics packages.
-
Replacing ff with Andromeda for storage of large data objects.
Bugfixes:
- Throw meaningful error when there are not enough negative controls with sufficient outcome count to perform positive control synthesis.
Changes:
-
Renamed synthesizePositiveControls to synthesizeReferenceSetPositiveControls, injectSignals to synthesizePositiveControls, and changed addExposureDaysToEnd to endAnchor.
-
Dropping individual metric function. The computeMetrics function now computes the standard set of metrics.
-
Added checkCohortSourceCodes and findOrphanSourceCodes functions.
Bugfixes:
- Fixed bug in positive control synthesis causing error when setting riskWindowStart to value other than 0.
Bugfixes:
- Fixed bug in positive control synthesis causing only most predictive covariates to be used when predicting outcome probabilities.