改变政策实施时间的安慰剂检验的结果图非常奇怪，和平常看到额安慰剂检验结果相差很多，求助这是为什么
代码如下：
use "F:\innovation.dta",clear

mat b = J(500,1,0)
mat se = J(500,1,0)
mat p = J(500,1,0)

forvalues i=1/500{
        use "F:innovation.dta", clear
        xtset daima year  
        drop if treat<1
        sample 1, count by(daima)   
        keep daima year
        rename year policy_year
        save match_id.dta, replace  
        merge 1:m daima using "F:innovation.dta"
        xtset daima year
        gen treat2 = (_merge == 3)
        gen period = (year >= policy_year)
        gen dd = treat2*period
        reg innovation dd lnpergdp third finance lnpeople i.daima i.year
        mat b[`i',1] = _b[dd]
        mat se[`i',1] = _se[dd]
        mat p[`i',1] = 2*ttail(e(df_r), abs(_b[dd]/_se[dd]))
}

生成的图片如下：
求助各位大佬，这是怎么回事呀？