请教各位大侠一个问题!
我通过一系列过程把一组不稳定的时间序列数据转化成稳定的数列。
这是我的do file 里面的命令
/*Plot the series*/
gen lp=log(ppi)
label variable lp "log of ppi"
tsline lp
ac lp
pac lp
/*step 1 determinate the lags*/
dfuller lp, trend lags(4) regress
estat ic
dfuller lp, trend lags(2) regress
estat ic
dfuller lp, trend lags(1) regress
estat ic
gen dlp=lp-l.lp
label variable dlp "differcial log of ppi"
regress dlp time l.lp l.dlp l2.dlp
predict e1,r
regress dlp time l.lp l.dlp
predict e2,r
label variable e1 "Residuals with two lags"
label variable e2 "Residuals with 1 lag"
corrgram e1, lag(20)
corrgram e2, lag(20)
/*step 2 test trend(t)*/
regress dlp time l.lp l.dlp l2.dlp
test time l.lp
/*step 3 Run ADF without deterministic trend*/
dfuller lp, lags(2) regress
/*step 4 test constant (drift term)*/
regress dlp l.lp l.dlp l2.dlp
test _cons l.lp
/*step 5 Run ADF without deterministic trend) and no drift term*/
dfuller lp, noconstant lags(2) regress
然后我确定了最后的转化方程是
Δyt =
βyt-1 +
β1 Δyt-1+β2 Δyt-2
这个是系数表
我的问题是,是不是要再通过那个系数和方程建立一列新的数列
然后那个数列才是稳定的?