TEMPORAL STRUCTURE · DIAGNOSTICS · FORECASTING · DECISION

Time Series Analysis

Diagnose how a measured process changes through time, compare forecasts against honest baselines, and return uncertainty with explicit decision boundaries.

POINT 01

Is this data a time series

CENTRAL QUESTION

Is this data a time series, or a table that happens to have dates in it?

PREDICT

Before moving anything, predict which control flips the state first.

Optional — commit to one expectation first, then experiment.

Ordered but irregular

Order matters; most time-series machinery needs a regular grid first.

Boundary: Irregular timestamps are ordered but not yet a series.

Explanatory simulation · bands are illustrative

EVIDENCE TO INSPECT

Three cases from exercise 1.3: daily maximum temperatures for a city over a year; temperatures taken at random moments; sales of 500 products in one month. Only the first is a time series with a usable frequency.

Inspect

Method
Threshold model on one numeric control; categorical inputs shift the bands.
Parameters
Bands and shifts stated in the figure.
Provenance
Illustrative values. Not derived from measurement.

Python code companion

Inspect the computational step

s = pd.Series(values, index=pd.date_range("2020-01-01", periods=6, freq="MS"))
s.index.freq          # None means pandas could not infer a frequency
s.index.is_monotonic_increasing

Illustrative code · inspect before execution

DAILY REVIEW

If you shuffled the rows and the analysis still worked, it was not time-series analysis.