如题，取对数后一阶差分结果：
> dlx=diff(log(x))
> adf.test(dlx)

        Augmented Dickey-Fuller Test
data:  dlx
Dickey-Fuller = -3.592, Lag order = 3, p-value = 0.04158
alternative hypothesis: stationary


p-value = 0.04158<0.05已经通过平稳性检验

二阶差分结果：
> dx=diff(log(x),diff=2)
> adf.test(dx)

        Augmented Dickey-Fuller Test
data:  dx
Dickey-Fuller = -5.3721, Lag order = 3, p-value = 0.01
alternative hypothesis: stationary

警告信息：
In adf.test(dx) : p-value smaller than printed p-value


像这种情况是在一阶差分的基础上进行模型选择还是二阶差分呢？为什么？