IPW Extreme Weight Truncation Threshold

The IPW Extreme Weight Truncation Threshold feature enables you to automatically remove observations with calculated weights that are above the specified threshold to avoid potential variance inflation.

If an observation has a low probability of being assigned to the treatment that it received according to the specified treatment model, then its inverse probability weight (IPW) could be extremely high. The variance of the casual effect estimate can increase when a few observations are weighted heavily.

Caution: Removing heavily weighted data points can bias the effect estimate if the observations are truly representative of the sample.