Investigation should be performed on whether a single imputation method for a single time series should be implemented opposed to the current switching strategy. The current methodology tests and select across five different methodology for each missing value, however this generates inconsistency overtime as the assumed data generation mechanism is not the same. Furthermore, it increases the relative error present in the data.
Investigation should be performed on whether a single imputation method for a single time series should be implemented opposed to the current switching strategy. The current methodology tests and select across five different methodology for each missing value, however this generates inconsistency overtime as the assumed data generation mechanism is not the same. Furthermore, it increases the relative error present in the data.