英文摘要
Conditionalaveragetreatmenteffect(CATE)isdesignedtocapturetheheterogeneityoftreatmenteffectacrosssubpopulations.Inthispaper,weproposeanewnonparametricestimationstrategyforCATEbasedonthepropensityscoreandprojectiontheory.Theproposedapproachhastwoadvantagesovertheexistingones.First,itdoesnotneedtoestimatethetwononparametricregressionfunctionsoftheout