From the course: Python for Time Series Forecasting
Join today to access over 24,800 courses taught by industry experts.
Data transformations to achieve stationarity - Python Tutorial
From the course: Python for Time Series Forecasting
Data transformations to achieve stationarity
- [Instructor] As a summary for this lesson on data stationarity assumptions, we have got the original data whose series wasn't stationarity, neither in the mean nor in the variance. At the moment we applied the logarithmic transformation, we see that the variance is constant as we move forward, but still the mean is not constant because here can be around five and then here could be around 5.8. It is not until we differentiate the time series that we get a constant mean. And also, because it was applied to the logarithmic data, we have the constant variance as well, achieving stationarity. If you see the bottom left, we have a constant mean, but not a constant variance, as the amplitudes around this area is very low compared to the one in the year '60. And if we apply the Dickey-Fuller tests on these data sets, we observe that the only one significant is the log applied transformation having been differentiated. So once we know we must work with the logarithmic data because the…
Download courses and learn on the go
Watch courses on your mobile device without an internet connection. Download courses using your iOS or Android LinkedIn Learning app.