1 / 24

Lecture 9 Forecasting Seasonal Data

Lecture 9 Forecasting Seasonal Data. Materials for this lecture Lecture 9 Seasonal Analysis.XLSX Read Chapter 15 pages 8-18 Read Chapter 16 Section 14. Uses for Seasonal Models. Have you noticed a difference in prices from one season to another? Tomatoes, avocados, grapes, lettuce

naguilar
Download Presentation

Lecture 9 Forecasting Seasonal Data

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. Lecture 9 Forecasting Seasonal Data • Materials for this lecture • Lecture 9 Seasonal Analysis.XLSX • Read Chapter 15 pages 8-18 • Read Chapter 16 Section 14

  2. Uses for Seasonal Models • Have you noticed a difference in prices from one season to another? • Tomatoes, avocados, grapes, lettuce • Wheat, corn, hay • 450-550 pound Steers • Gasoline • Many price series have seasonal patterns • To forecast seasonal data you must explicitly incorporate variables in the model to pick up the seasonality

  3. Seasonal and Moving Average Forecasts • Monthly, weekly and quarterly data generally have a seasonal pattern • Annual data have cycles • Seasonal patterns repeat each year due to • Seasonal production due to climate or weather • Seasonal demand (holidays, summer, etc.) • Cycle may also be present and the seasonal pattern is mapped over the top of the cycle if you are using many years of monthly data

  4. Seasonal and Moving Average Forecasts • Example of monthly prices showing seasonal variability on top of a multi year cycle

  5. Econometric Models for Forecasting Seasonal Patterns • Seasonal indices • Composite forecast models • Dummy variable regression model • Harmonic regression model • Moving average model

  6. Steps for Estimating a Seasonal Index • Graph the data • Check for a trend and seasonal pattern • Develop and use a seasonal index if no trend is present • If a trend is present, forecast the trend and combine it with a seasonal index • Develop a composite forecast model • Trend and seasonal components

  7. Two kinds of Seasonal Indices • Price Index • The traditional index value shows the relative relationship of price between months or quarters • It is ONLY used with price data • Fractional Contribution Index • If the variable is a quantity, calculate a fractional contribution index to show the relative contribution of each month to the annual total quantity • It is ONLY used with quantities

  8. Seasonal Index Model • Seasonal index is a simple way to forecast a monthly or quarterly data series • Index represents the fraction that each month’s price or sales is above or below the annual mean • If the seasonal index for June is 1.10 it means the June price averages 110% of the annual average price • If the seasonal index for December is 0.85 it means the December price averages 85% of the annual average price • A function in Simetar estimates the seasonal index for time series with a few simple steps, so it is very easy to use this type of forecasting model

  9. Seasonal Index Model Estimated by Simetar

  10. Using a Seasonal Price Index for Forecasting • Seasonal index has an average of 1.0 • Use seasonal index to forecast monthly prices from annual average price forecast PJan = Annual Avg Price * IndexJan PMar = Annual Avg Price * IndexMar • Assume you have a trend forecasted annual average price of $125, you can develop monthly forecasts using the seasonal price indices, as: Jan Price = $125 * 0.600 = $75.0 Mar Price = $125 * 0.976 = $122.0 • The forecast can include risk by using a stochastic forecast of annual price

  11. Seasonal Fractional Contribution Index • Fractional Contribution Index sums to 1.0 to represent 100% of annual quantity of sales • Each month’s value is the fraction of total sales that month • Use a trend or structural model to forecast annual sales SalesJan= Total Annual Sales * IndexJan SalesJun= Total Annual Sales * IndexJun • For an annual sales forecast at 340,000 units SalesJan= 340,000 * 0.050 = 17,000.0 SalesJun= 340,000 * 0.076 = 25,840.0 • This forecast is useful for input procurement and inventory management • The forecast can include risk by using a stochastic forecast of annual sales

  12. Seasonal Forecast Using Dummy Variable Models • Dummy variable regression model can forecast trend and seasonal varibility • Include a trend if one is present • Regression model to be estimated is: Ŷ = a + b1Jan + b2Feb+…+b11Nov + b13T + b14Z • Jan – Nov are individual dummy variable 0’s and 1’s • Effect of Dec is captured in the intercept • If the data are quarterly, use 3 dummy variables, for first 3 quarters and intercept picks up affect for fourth quarter Ŷ = a + b1Qt1 + b2Qt2 + b11Qt3 + b13T+ b14Z • NOTE: the Z variable represents other structural variables that can be included in the model

  13. What is Each Part of the Equation Doing? Ŷ = a + b1Jan + b2Feb+…+b11Nov + b13T + b14Z • The “a” intercept is capturing the average value of Y when all the other variables are zero • The b1 through b11 are capturing the effects of different months on the Y variable, some may be positive and others negative. Some may not be statistically significant but should be left in the model for completeness • The “T” variable is trend and effectively “de-trends” the data prior to estimating the seasonal effects • The “Z” variable represents other structural variables that can be included in the model, such as income, population, own price and competing prices

  14. Seasonal Forecast with Dummy Variable Models • Set up the X matrix with 0’s and 1’s • Easy to forecast as the seasonal effect is assumed to persist forever • Note the pattern of 0s and 1s for months • December affect is captured in the intercept

  15. Seasonal Forecast with Dummy Variable Models • Regression Results for a Monthly Dummy Variable Model • May not have a statistically significant effect for each month, as indicate by the Student t on the betas • Monthly forecasts use beta for the month being forecasted • Jan forecast = 45.93+4.147 * (1) +1.553*T -0.017 *T2 +0.0001 * T3 • May forecast = 45.93 + 4.999 * (1) +1.553*T -0.017 *T2 +0.0001 * T3

  16. Probabilistic Monthly Forecasts • In this case we use the Adjusted Stochastic Indices at the bottom of the Simetar output

  17. Probabilistic Monthly Forecasts • Develop probabilistic forecast of the annual price • In this case I used a trend forecast of the annual average prices and forecasted the value for year 18 • Use the stochastic Indices to simulate stochastic monthly forecasts

  18. Harmonic Regression for Seasonal Models • Add Sin and Cos functions in OLS regression to isolate seasonal variation • Define a variable SL to represent alternative seasonal lengths: 2, 3, 4, … • Create the X Matrix for OLS regression X1 = Trend so it is: T = 1, 2, 3, 4, 5, … . X2 = Sin(2 * ρi() * T / SL) X3 = Cos(2 * ρi() * T / SL) Fit the regression equation of: Ŷi = a + b1T + b2 Sin((2 * ρi() * T) / SL) + b3 Cos((2 * ρi() * T) / SL) + b4T2 + b5T3 • Only include “b1T” if a trend is present

  19. Harmonic Regression for Seasonal Models This is what the X matrix looks like for a Harmonic Regression

  20. Harmonic Regression for Seasonal Models • Note that the COS term is not statistically significant, but it MUST be included when using the model to forecast • SIN and COS are creating the wave effect in the forecast • T is creating the positive trend in the forecast • The model needs more terms to capture the underlying cycle

  21. Moving Average Forecasts • Moving average forecasts are used by the industry as the naive forecast • If you can not beat the MA then you can be replaced by a simple MA forecast • Calculate a MA of length K periods and move the average each period, drop the oldest and add the newest value 3 Period MA Ŷ4 = (Y1 + Y2 + Y3) / 3 Ŷ5 = (Y2 + Y3 + Y4) / 3 Ŷ6 = (Y3 + Y4 + Y5) / 3

  22. Moving Average Forecasts • Example of a 12 Month MA model estimated and forecasted with Simetar • Change Periods value to test alternative MA lengths • MA with lowest MAPE is best but still leave a couple of periods, 3 is probably the minimum

  23. Probabilistic Moving Average Forecasts • Use the MA model with lowest MAPE but with a reasonable number of periods (min of 3) • If the no. periods is to short soon you are using forecasted values to forecast more forecasted values • Assume your last real value is denoted as T, then a forecast for period T+5 is using all forecasted values given a 4 period moving average ỸT+5 = (YT+1 +YT+2 + YT+3 + YT+4)/4

  24. Probabilistic Forecast for Seasonal Models • Stochastic simulation used to develop a probabilistic forecast for a random variable Ỹi = NORM(Ŷi , σ) Ŷi is the forecast for each future period based on your expectations of the exogenous variables for future periods σ is a constant and is calculated as the stddev of the residuals

More Related