(See also #52 on Gaussian Processes for outlier detection.)
I'm evaluating different strategies for automated Bayesian outlier detection, but it would be very helpful to have an idea of the physical mechanism causing outliers since this will affect how we deal with it in the code.
Some possibilities would affect all measurements:
Erroneous compound dispensing via the HP D300
Erroneous protein dispensing via the EVO
Incomplete mixing
Ligand or protein aggregation
Some effects might affect only one type of measurement:
Dust particles might only affect top or bottom reads
Scratch, smear, or defect in plate may only affect bottom reads
Any other ideas?
@sonyahanson has prepared a nice dataset that illustrates this here, but it may need to be updated to the new scheme for representing input for quickmodel.
(See also #52 on Gaussian Processes for outlier detection.)
I'm evaluating different strategies for automated Bayesian outlier detection, but it would be very helpful to have an idea of the physical mechanism causing outliers since this will affect how we deal with it in the code.
Some possibilities would affect all measurements:
Some effects might affect only one type of measurement:
Any other ideas?
@sonyahanson has prepared a nice dataset that illustrates this here, but it may need to be updated to the new scheme for representing input for quickmodel.