A brief note on Time Series and Moving Averages . Read it before a test. Try each question at the end before you open its answer.
The steps Plot the values against time and join the points. Moving averages over one full cycle: 3 points for terms, 4 for quarters, 7 for days, 12 for months. Centre them when the number of points is even. Trend line by eye through the moving averages. Seasonal variation = = = actual − - − trend. Seasonal component = = = the mean of the variations of one season. Forecast = = = trend value + + + seasonal component. The ideas Trend, seasonal variation and random variation are the three movements. An odd-point average sits on the middle time. An even-point average sits between two times. The components should add up to about zero. A forecast assumes the trend and the pattern carry on. It is less reliable further ahead. A worked example A kiosk sells ice creams, in hundreds, each quarter: year 1 12, 30, 44, 14; year 2 16, 34, 48, 18.
4-point averages. 25, 26, 27, 28 and 29. Centred averages. 25.5, 26.5, 27.5 and 28.5, at quarters 3 to 6. A seasonal variation. At quarter 3: 44 − 25.5 = 18.5 44 - 25.5 = 18.5 44 − 25.5 = 18.5 hundred. Practice with new numbers
A program to try The program works out the averages of the worked example.
Run it. Which line prints the variation of step 3?
1. Add the variation of quarter 4: sales[3] - centred[1]. Is it positive or negative?
Mistakes to avoid Centring an odd-point average. Only even-point averages need it. Variation the wrong way round. Actual minus trend. Dropping the sign of a component in a forecast. Check yourself Find the 3-point moving average of 21, 27 and 36. Stuck? Show a hint Show the answerCentre the 4-point averages 40 and 43. Stuck? Show a hint Show the answerThe trend value is 96 and the component is − 7.5 -7.5 − 7.5 . Give the forecast. Stuck? Show a hint Show the answerWhy is a forecast for far ahead less reliable? Stuck? Show a hint Show the answer