Data Officer · Synthetic data demonstration

Exploring a
forecasting approach.

My historical Python study used public Census data to explore exponential smoothing. Here, invented index values demonstrate the approach.

Historical notebook: 24 March 2023Synthetic data · sample periods 1–8
Forecasting demonstration with synthetic dataInvented index inputs of 74, 81, 76, 89 and 85 span sample periods 1 to 5. A fixed-parameter multiplicative-trend smoothing calculation produces a synthetic projection for periods 6 to 8. No observed farm counts, real years or historical model outputs are shown. Forecasting demonstration · synthetic dataIllustrative index · sample periods 1–8 708090100110 85 105.5 12345678 Invented inputsSmoothed levelIllustrative projection
All inputs are invented. A simple multiplicative-trend smoothing calculation generates the projection; the shaded area shows the next three sample periods.
01 / SELECT

Start with a source

Retrieve Statistics Canada’s livestock table, select Canada and dairy cows, and isolate the number of farms reporting.

02 / PREPARE

Make time explicit

Index the observations by year, interpolate an annual series and separate a training period from a comparison period.

03 / EXAMINE

Fit and compare

Explore exponential smoothing with a multiplicative trend, plot the comparison and calculate error measures before extending the series.

The steps describe the historical study. The chart is a new synthetic example: its index values, sample periods and projection are not farm counts or estimates for any place or year.

Historical study source, not chart data: Statistics Canada, Table 32-10-0155-01 — Selected livestock and poultry, Census of Agriculture historical data.