I have a dataset of stock prices and wanted to make it stationary.

I did a difference using lag 1 and then did the difference again using lag 1.

After the first differencing the Augmented Dickey Fuller test for unit root had a p-value of 0.0001, yet the ACF had a sinusoid - I feel like this is telling me that my differences (my volatilty) have a periodicity, my variance is non-stationary - is this a correct deduction? How come the sinusoid appears but the DF test still has a low p-value?

The ACF for the raw data was a negative gradient straight line. So did the 1st order difference create or did it expose seasonality/periodicity in the data (seasonality in the variance)?

EDIT : I have just read that cyclic behaviour does not equal seasonality. So potentially even with the cycles in the ACF - the 1st order difference can still be classed as stationary?

Finally: Can I look at the absolute value of the differences to avoid negative values, when I come to build my GARCH model? Can I remove the MA from my dataset to make it stationary?

Answers to my questions appreciated!

Raw Data - No Differencing Raw Data - No Differencing After 1st differencing After 1st differencing After 2nd differencing After 2nd differencing


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.