Hello
There are two types of negative controls: outcomes and predictors.
Atlas population estimation has a section “Negative Control Outcome Cohort Definition”.
Atlas help explains negative controls: “Negative control outcomes are outcomes that are not believed to be caused by either the target or the comparator, and where therefore the true hazard ratio equals 1”.
But EmpiricalCalibration, the R package used to perform empirical calibration of observational study, estimates “using a set of negative control hypotheses”. “Negative controls are test-hypotheses where the exposure is not believed to cause the outcome”.
The EmpiricalPCalibration vignette shows an example using predictor controls.
The MultipleAnalysis vignette of the CohortMethod package also uses predictor controls.
Then I’m confused.
In Atlas, what kind of controls do we need to specify, predictors or outcomes?
Or Atlas doesn’t use the EmpiricalCalibration package?
Regards