This function implements an existing model
evaluateExistingModel(modelTable, covariateTable, interceptTable = NULL, type = "score", covariateSettings, customCovariates = NULL, addExposureDaysToStart = F, riskWindowStart = 1, addExposureDaysToEnd = F, riskWindowEnd = 365, requireTimeAtRisk = T, minTimeAtRisk = 364, includeAllOutcomes = T, removeSubjectsWithPriorOutcome = T, priorOutcomeLookback = 99999, verbosity = "INFO", washoutPeriod = 0, firstExposureOnly = F, binary = T, connectionDetails, cdmDatabaseSchema, cohortDatabaseSchema, cohortTable, cohortId, outcomeDatabaseSchema, outcomeTable, outcomeId, oracleTempSchema = cdmDatabaseSchema, modelName = "existingModel", calibrationPopulation = NULL, covariateSummary = T, cdmVersion = 5)
modelTable | The model covariates and scores |
---|---|
covariateTable | The mapping from model covariates to standard covariates |
interceptTable | The model intercepts |
type | Model type (score or logistic) |
covariateSettings | The standard covariate settings (specify covariate lookback time) |
customCovariates | A table of covariateId, sql (sql creates the custom covariate) |
addExposureDaysToStart | riskWindowStart relative to the cohort end date instead of the cohort start date? |
riskWindowStart | The day after index to start predicting the outcome |
addExposureDaysToEnd | riskWindowEnd relative to the cohort end date instead of the cohort start date? |
riskWindowEnd | The day after index to stop predicting the outcome |
requireTimeAtRisk | Do you want to ignore people who leave the database some point between the riskWindowStart and riskWindowEnd |
minTimeAtRisk | If requireTimeAtRisk is TRUE, how many days must they be observed before leaving to get included (default recommendation is all risk period: riskWindowEnd-riskWindowStart) |
includeAllOutcomes | Setting this to TRUE means people with the outcome who leave the data during the risk period are still included, so only non-outcome people who leave during the risk period are removed |
removeSubjectsWithPriorOutcome | Remove people from the target population if they have the outcome prior to target cohort start date |
priorOutcomeLookback | Lookback for removeSubjectsWithPriorOutcome |
verbosity | The study population creation verbosity |
washoutPeriod | Remove patients from the population with less than washoutPeriod of days prior observation |
firstExposureOnly | If patients are in the target population multiple times, use only the first date |
binary | Binary classificsation (T or F) |
connectionDetails | The details to connect to the CDM |
cdmDatabaseSchema | A string specifying the database containing the cdm |
cohortDatabaseSchema | A string specifying the database containing the target cohorts |
cohortTable | A string specifying the table containing the target cohorts |
cohortId | An iteger specifying the cohort id for the target cohorts |
outcomeDatabaseSchema | A string specifying the database containing the outcome cohorts |
outcomeTable | A string specifying the table containing the outcome cohorts |
outcomeId | An iteger specifying the cohort id for the outcome cohorts |
oracleTempSchema | The temp oracle schema |
modelName | The name of the model |
calibrationPopulation | A data.frame of subjectId, cohortStartDate, indexes used to recalibrate the model on new data |
covariateSummary | Whether to calculate the covariateSummary |
cdmVersion | The CDM version being used |
The performance of the existing model and prediction
Implements an existing model and evaluates its performance