Can it change after publication?

Yes, estimated components can be revised.

Does seasonal adjustment remove it?

Generally no; the adjusted series retains it.

AI Glossary / Statistics and probability / Advanced

Irregular component of a time series

Definition

The irregular component of a time series is the remainder after estimated trend-cycle, seasonal, and relevant calendar effects have been removed. It represents movements not assigned to those components by the chosen decomposition. Calling a movement irregular does not, by itself, establish its cause or prove that it is independent random noise.

What it means

A monthly business measure can rise because its underlying level is increasing, because that month is usually busy, or because something unusual occurred. Decomposition tries to distinguish these influences.
The irregular component is estimated rather than directly observed. Its values depend on how the other components are estimated and on the information available at the time. A new observation can change an earlier decomposition even when the earlier recorded total remains unchanged.

How it works

An additive representation writes the observed value as trend-cycle plus seasonal and calendar effects plus irregular. Subtraction then gives an irregular value in the original measurement units.
A multiplicative representation uses a product of components. Dividing the observed value by the other components produces an irregular factor, with one indicating no irregular departure. This is useful when fluctuations are proportional to the series level, provided the data and model support it.
A decomposition remainder need not be white noise. Filter-based decompositions can leave correlation in their irregular components. Check the method's assumptions and examine remaining patterns before interpreting a residual as unpredictable or using it to set an alert threshold.

Three practical examples

An AI support team separates predictable monthly demand from departures that may warrant investigation, such as an unexpected outage-related surge.
A retailer compares seasonally adjusted sales with an estimated trend to understand why one month's movement differs from the broader direction.
A forecasting analyst checks whether a decomposition remainder still contains a recurring calendar pattern, which could indicate incomplete adjustment.

Think of it this way

Think of a recorded tune with a steady underlying melody and a repeating rhythm. After estimating and removing those parts, the remaining sound is the remainder. It may contain accidental noise, but it may also contain a musical feature the separation process failed to recognize.

Worked example

Consider an original additive illustration measured in hundreds of support requests. Four consecutive observations are 121, 115, 119, and 125. Suppose a decomposition supplies trend-cycle estimates of 100, 102, 104, and 106, and combined seasonal/calendar effects of 20, 15, 10, and 15.
Subtracting both estimated components gives irregular values of 1, minus 2, 5, and 4. For the third observation, 119 minus 104 minus 10 = 5. The reconstruction is exact: 104 plus 10 plus 5 = 119.
Removing only the seasonal/calendar effects gives adjusted observations of 101, 100, 109, and 110. These still contain the irregular component. From the second to the third observation, the adjusted series rises by nine, although the estimated trend-cycle rises by only two.
The remaining seven units of that adjusted movement come from the irregular change: five minus minus two = seven. Thus a nine-unit increase in the adjusted series should not automatically be described as a nine-unit strengthening of the underlying trend.
Now imagine that additional data lead the analyst to revise the third period's trend-cycle from 104 to 106, while retaining its observed total of 119 and seasonal/calendar effect of 10. Its estimated irregular component becomes three. Nothing about the recorded observation changed; its allocation across estimated components did.
The first four irregular values have mean two, since their sum is eight. A short segment's estimated remainder need not have a zero sample mean simply because a statistical model assumes a zero-mean disturbance over repeated realizations.
For a separate multiplicative illustration, take an observed value of 132, a trend-cycle of 100, and a seasonal/calendar factor of 1.2. The irregular factor is 132 divided by 120 = 1.1. The adjusted value is 132 divided by 1.2 = 110, ten percent above the trend-cycle.
If the next observation is 129.6 with trend-cycle 120 and the same seasonal factor 1.2, its irregular factor is 0.9 and its adjusted value is 108. The trend-cycle rose by twenty percent, but the adjusted value fell from 110 to 108 because the irregular factor changed substantially.
An irregular factor of 1.1 is not an additive difference of 1.1 requests. In the first multiplicative example, it corresponds to twelve extra units relative to the unadjusted trend-times-seasonal baseline of 120, or ten units on the seasonally adjusted scale.
These supplied components demonstrate arithmetic and interpretation. Four observations would not be adequate evidence for estimating a stable annual seasonal pattern; a real analysis needs an appropriate history and diagnostic checks.

Illustrative case study

Fictional company Cedar Support initially treats every increase in adjusted request volume as sustained growth. Its analyst separates the adjusted series into an estimated trend-cycle and an irregular remainder.
A large positive remainder prompts an operational investigation rather than an immediate permanent staffing increase. The team preserves the original observation, records any explanatory evidence, and tracks revisions as more periods arrive. This is a fictional decision process, not evidence that every spike has a known explanation.

Common misconceptions and mistakes

Seasonally adjusted does not mean irregular-free.
Irregular does not mean incorrectly measured or safe to delete.
A large remainder does not prove a particular event caused it.
Additive residuals and multiplicative factors use different scales.
A remainder with persistent structure may indicate an inadequate decomposition rather than inherently unpredictable behavior.

Business use

The irregular component can help identify periods requiring investigation and distinguish short-term movements from sustained changes. Present it alongside the original series and decomposition assumptions. For AI forecasting, evaluate methods using only information available at each historical forecast date; revised full-history components can otherwise create an unrealistically favorable evaluation.

Check your understanding

What is the third additive irregular value?

Answer: Five.

How much of the adjusted nine-unit rise came from its change?

Answer: Seven.

What does a multiplicative factor of one indicate?

Answer: No irregular departure from the other components' product.

Related terms

Seasonal adjustment
Residual
White noise
Time series decomposition
trend-cycle
calendar effect
forecast error
data revision

Learn next

Compare additive and multiplicative decompositions, then examine residual autocorrelation, endpoint revisions, and evaluation using historical data vintages.

Residual

Sources

U.S. Census Bureau, Time Series and Seasonal Adjustment Glossary: https://www.census.gov/topics/research/seasonal-adjustment/glossary.html . Australian Bureau of Statistics, Time Series Analysis: The Process of Seasonal Adjustment: https://www.abs.gov.au/websitedbs/d3310114.nsf/51c9a3d36edfd0dfca256acb00118404/5fc845406def2c3dca256ce100188f8e!OpenDocument . Numerical illustrations and company scenario are original.

Previous: Iris recognition Next: Irregular effect