# ARIMA(0,1,0) random walk with drift

#### hlsmith

##### Less is more. Stay pure. Stay poor.
What code do I use to fit the above model, or what type of differencing do I do?

Something like a lag 1 subtraction?

My model seemed like it needed differencing per plots and the auto.arima came back with the above structure.

Thanks.

#### GretaGarbo

##### Human
As I understand the above there are no AR-parameters and no MA-parameters so there is nothing to "fit". It just needs to take the first difference to get a more stationary series. I believe that you need to declare the data as a time series.

So, something like:

Code:
x_new <- x - lag(x, 1)  # take the first lag

#### hlsmith

##### Less is more. Stay pure. Stay poor.
Greta, Thank you that is how I was proceeding.

Hmm, I am still a time series novice. So this definitely makes my data stationary, though I don't know what to do with it now? I was comparing a serie before and after an intervention. So I was looking at the coefficient for the initial slope then how much it increased after intervention followed by any change in slope trend. The differencing appears to have removed my jump up since I differenced each of the two groups independently. Any suggestions, I know this is a different questions then initially posted?

#### Miner

##### TS Contributor
hlsmith,

I can think of two possible approaches if both sets are stationary after differencing.

The first would be to plot both sets of differenced data on a single individuals control chart. The control limits should be calculated using the first set of differenced data, so any changes between the two sets will appear as a run. Any changes in slope should appear as a change in the mean difference, and should produce a run. No run means no change in slope.

Otherwise, try a 2-sample t-test on the differences.

#### hlsmith

##### Less is more. Stay pure. Stay poor.
Thanks for your input, I will review them tonight.