Read the pattern in ordered data, then project it forward
Time series analysis is the study of data recorded in order — how observations depend on the ones before them, and how that dependence can be turned into a description and a forecast. It runs from the components a series is made of (trend, seasonality, noise), through stationarity and the autocorrelation function, into linear time-domain models, the frequency-domain view, nonstationary and seasonal models, volatility models for financial data, and state-space form. Nano College takes the standard university treatment of this subject and rebuilds it as short, self-contained pieces: each article can be read and finished on its own, with the rigour kept and the padding removed. Working through them leaves you able to choose a model class, estimate it, check whether it survived contact with the data, and attach a defensible error to the forecast.
Coming soon
Establish what a series is made of before modelling any of it.
Establish what a series is made of before modelling any of it.
The time-domain linear models, and what their correlation patterns reveal.
The time-domain linear models, and what their correlation patterns reveal.
The same series seen as a mixture of cycles at different frequencies.
The same series seen as a mixture of cycles at different frequencies.
Turning a candidate order into a model you can defend.
Turning a candidate order into a model you can defend.
What to do when the series will not hold still.
What to do when the series will not hold still.
Why the variance moves, and how to model it.
Why the variance moves, and how to model it.
When one series is not enough, and when the hidden state is the point.
When one series is not enough, and when the hidden state is the point.
Ready-made forecasters, and the assumptions worth suspecting.
Ready-made forecasters, and the assumptions worth suspecting.
“A patient, complete route from autocorrelation to state-space forecasting.”
This collection keeps the rigour of a full course while cutting the padding written to fill a term. Each article stands alone, so you can start from the question you actually have: a stubborn seasonal pattern, a volatility spike, or a forecast that keeps missing. Problem sets and software tutorials are left out.
Nano College takes the standard university textbook for a course and turns it into short, self-contained pieces you can actually finish. We keep the ground the textbook covers and the rigour it insists on; what we leave out is the padding that only exists to fill a semester.