Converting non-stationary to stationary

问题

I have one data it is not stationary. I'm trying to make it stationary. I tried log transformation, BoxCox transformation, lag(1, 2 and 3) differences. No use of these transformations and differencing.

I used adf test to test stationary in R. Can anybody tell is there any other method to make it stationary.

data:

6.668
5.591
4.734
3.493
3.235
3.968
2.64
2.885
3.045
3.579
5.463
5.458
5.758
5.931
5.731
6.799
9.568
9.11
6.571
8.528
15.11
13.956
16.46
19.599
27.281
39.928
56.284
67.565
106.399
104.229
100.686
141.755
164.447

回答1:

You did not respond to my comment, so to make the point, if you try

> require(tseries)
> adf.test(diff(diff(log(data))))

then you get the response

        Augmented Dickey-Fuller Test

data:  diff(diff(log(data))) 
Dickey-Fuller = -5.1371, Lag order = 3, p-value = 0.01
alternative hypothesis: stationary 

Warning message:
In adf.test(diff(diff(log(data)))) : p-value smaller than printed p-value

来源：https://stackoverflow.com/questions/17192428/converting-non-stationary-to-stationary