想请教各位老师一个问题，我在处理面板数据的过程中，对数据进行了lag和winsor，然后进行xtreg之后，观测值大幅度降低（相比于summarize的时候），这样估计出来的结果就都不显著，想请问下是为什么呢？（我猜想是什么地方观测值本不应该但被omit了）。
附代码如下：
use "C:\Users\Administrator\Desktop\STUDY\thesis\Regression sample.dta"
drop if SHORT_TERM1 > 1
drop if SHORT_TERM2 > 1
drop if SHORT_TERM3 > 1
drop if SHORT_TERM4 > 1
drop if SHORT_TERM5 > 1
duplicates drop fyear gvkey ,force
xtset gvkey fyear
// z_scoredummy_lag_w
gen z_score_dummy = ( z_score >1.81)
gen z_score_dummy_lag = L.z_score_dummy
// creditratingdummy_lag_w
merge m:m year TICKER using "C:\Users\Administrator\Desktop\STUDY\thesis\credit ratings(mod).dta"
drop if _merge==2
drop _merge
duplicates drop fyear gvkey ,force
gen ratingdummy = ( splticrm!="")
xtset gvkey fyear
gen ratingdummy_lag = L.ratingdummy
// abnormal earnings
merge m:m fyear gvkey using "C:\Users\Administrator\Desktop\STUDY\thesis\EPS.dta"
keep if _merge==3
drop _merge
duplicates drop fyear gvkey ,force
xtset gvkey fyear
gen depspi = F.epspi - epspi
gen abnearn = depspi/ prcc_f
gen abnearn_lag = L.abnearn
// Motivation
gen size2 = size * size
gen size2_lag = L.size2

winsor2 abnearn_lag size2_lag

summarize SHORT_TERM1 SHORT_TERM2 SHORT_TERM3 SHORT_TERM4 SHORT_TERM5 z_score_dummy_lag ratingdummy_lag cf_volatility_lag_w asset_maturity_lag_w termspread_lag_w mktleverage_lag_w MtoB_lag_w size_lag_w abnearn_lag_w

// Regression
gen envienfor = Nor_Facilities * environment_score
gen envienfor_lag = L.envienfor
gen enforcement_lag = L.Nor_Facilities
gen environment_score_lag = L.environment_score

winsor2 enforcement_lag environment_score_lag envienfor_lag

xtreg SHORT_TERM1 size_lag_w MtoB_lag_w z_score_lag_w mktleverage_lag_w termspread_lag_w asset_maturity_lag_w ratingdummy_lag cf_volatility_lag_w environment_score_lag enforcement_lag envienfor size2_lag i.fyear, fe r

感谢解答！