THIS MEANS THE GENDER VARIABLE IS A MIXTURE VARIABLE WHICH CAN YIELD MIXTURE EFFECT, HENCE YOU CAN EXPLAIN THAT THE GENDER IS QUITE IMPORTANT FOR THE Y VARIABLE. ALSO YOU CAN USE COVARIANCE METHOD TO SOLVE YOU PROBLEM
原因应该很多。建议作全样本散点图看看，有个直观认识。
说不定两组gender形成两簇散点团，这就很有可能导致回归结果全样本显著，而分组后不显著。如：
clear
set seed 123456
set obs 1500
gen gender = (uniform()<=2/3)
gen x1 = 2 - 2*gender + rnormal()
gen y = x1 + 50*rnormal()
reg y x1
reg y x1 if gender
reg y x1 if !gender
sc y x1 if gender || sc y x1 if !gender
扫码加好友，拉您进群
全部版块
我的主页
收藏
