liveBook · Manning

Long-term trend

As explained in Grokking Bayes

A temporal trend is a systematic long-term direction of change in a time series.

A long-term trend is a systematic direction of change across time. Averaging monthly CO₂ measurements by year makes the persistent upward movement easier to see without seasonal variation. section 9.1

Mauna Loa atmospheric CO₂

Yearly average atmospheric CO₂ measured at Mauna Loa from 1958 to 2001 shows a clear upward trend. Forecasting from an earlier year, such as 1975, requires using only observations available then and accounting for dependence between nearby years. section 9.1

A filter that assumes the latent state’s expected value stays constant produces flat long-range mean predictions. Such a model reacts to new observations but has no mechanism for persistent drift, making it inadequate for strongly trending processes. section 9.10

Figure 9.1 Yearly average CO₂ concentrations measured at Mauna Loa Observatory between 1958 and 2001. The time series shows an upward trend.

The time-series plot makes the upward direction of yearly CO₂ concentrations immediately visible. section 9.1

How Grokking Bayes covers it

How deep it goes
Defines the pattern, demonstrates it with CO₂ data, and identifies persistent drift as missing from a random-walk filter.
Examples
Uses yearly Mauna Loa CO₂ measurements, a 1975 forecasting cutoff, and flat multi-year predictions from a random-walk filter.
What you can do after
Evaluate whether a forecasting filter can represent persistent drift rather than merely react to incoming observations.
How it connects
Connects long-term trend with time series, seasonal variation, forecasting, latent states, and state-space models.

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Grokking Bayes. The section links point to the parts of the book it draws on.