breakdown of data components

Original Data

Trend

Seasonal Components

Noise

Special Week Components

Underlying

services we provideStatistical Modelling, Forecasting & Seasonal Adjustment of Data

Breakdown
When data appears noisy, it is difficult to properly analyse the underlying movements. A dataset can be broken into its component parts using a variety of techniques to enable analysis of:

• Trend
• Seasonalilty
• Smoothed Fluctuations
• Special Cases
• Noise

In the example on the left, the original data appears noisy and difficult to interpret. By breaking the information into several component parts, it is possible to uncover the various significant items and interpret them individually.

Trend
This is seen as only a small (but important) part, growing at 1.7% p.a.

Seasonality
Sales in this example follow a yearly seasonality where volumes in the latter part of the year are far stronger than in the beginning.

Fluctuations
Once the other components are understood, a smoothing function (such as Tukey’s 4253H resistant smoother) can be applied to identify the other major movements within the data. By seeing these clearly and attempting to match changes with known causes, better planning for future events can be undertaken.