Forecasting using SAS Forecast Server, SAS/ETS, and more

Time series stationarity Dickey Fuller

Reply
New Contributor
Posts: 4

Time series stationarity Dickey Fuller

[ Edited ]

I have a time series with prices.  And i need to check if it stationary or not ?And if not, i need to make it stationary.

 

So i think, i have to do a Dickey Fuller test.

 

So i tried this code

proc ARIMA data=prices;
identify var=price stationarity=(adf=(0)) ;
run;

 

Is it correct ?

 

Super User
Posts: 9,671

Re: Time series stationarity Dickey Fuller

ACF is decreasing very slowly, therefore price could not be stationarity.

Difference it .

 

identify var=price(1)
New Contributor
Posts: 4

Re: Time series stationarity Dickey Fuller

[ Edited ]

i see that's the same like:

newprice=price-lag1(price);

 

It is stationary now, but some prices are negative.

Super User
Posts: 9,671

Re: Time series stationarity Dickey Fuller

Use SAS syntax (forecast statement), and SAS will take care of it.

identify var=price(1);
estimate p=2;
forecast lead=12 


Super User
Posts: 9,671

Re: Time series stationarity Dickey Fuller

If you want constraint price be positive , Try PROC ESM .

Or "Forecasting Log Transformed Data" , check it in PROC ARIMA's documentation.


New Contributor
Posts: 4

Re: Time series stationarity Dickey Fuller

I have very small variables. It's gas prices.

So after difference i get for example some prices like -0.0568, -0.2654

I tried this

data new;
set newprices;
ylog = log( newprices );
run;

But  after this i get :

-2.356,-2.0325  etc.

 

i don't understand

Super User
Posts: 9,671

Re: Time series stationarity Dickey Fuller

Sorry . mislead you.

 

newprice=price-lag1(price);

 

are supposed to get negative value like -0.0568, -0.2654.

if you put model on newprice ,you will get difference of price 's forecast,

therefore suggest you to use PROC ARIMA syntax, not calculated it by hand .

Ask a Question
Discussion stats
  • 6 replies
  • 166 views
  • 2 likes
  • 2 in conversation