smooth |

Exponential smoothing.

Forecasts a series using one of a number of exponential smoothing techniques. By default, smooth estimates the damping parameters of the smoothing model to minimize the sum of squared forecast errors, but you may specify your own values for the damping parameters.

smooth automatically calculates in-sample forecast errors and puts them into the series RESID.

Syntax

series_name.smooth(method) smooth_name [freq]

You should follow the smooth keyword with a name for the smoothed series. You must also specify the smoothing method in parentheses. The optional freq may be used to override the default for the number of periods in the seasonal cycle. By default, this value is set to the workfile frequency (e.g. — 4 for quarterly data). For undated data, the default is 5.

Options

Smoothing method options

s[,x] | Single exponential smoothing for series with no trend. You may optionally specify a number x between zero and one for the mean parameter. |

d[,x] | Double exponential smoothing for series with a trend. You may optionally specify a number x between zero and one for the mean parameter. |

n[,x,y] | Holt-Winters without seasonal component. You may optionally specify numbers x and y between zero and one for the mean and trend parameters, respectively. |

a[,x,y,z] | Holt-Winters with additive seasonal component. You may optionally specify numbers x, y, and z, between zero and one for the mean, trend, and seasonal parameters, respectively. |

m[,x,y,z] | Holt-Winters with multiplicative seasonal component. You may optionally specify numbers x, y, and z, between zero and one for the mean, trend, and seasonal parameters, respectively. |

Other Options:

prompt | Force the dialog to appear from within a program. |

p | Print a table of forecast statistics. |

If you wish to set only some of the damping parameters and let EViews estimate the other parameters, enter the letter “e” where you wish the parameter to be estimated.

If the number of seasons is different from the frequency of the workfile (an unusual case that arises primarily if you are using an undated workfile for data that are not monthly or quarterly), you should enter the number of seasons after the smoothed series name. This optional input will have no effect on forecasts without seasonal components.

Examples

sales.smooth(s) sales_f

smooths the SALES series by a single exponential smoothing method and saves the smoothed series as SALES_F. EViews estimates the damping (smoothing) parameter and displays it with other forecast statistics in the SALES series window.

tb3.smooth(n,e,.3) tb3_hw

smooths the TB3 series by a Holt-Winters no seasonal method and saves the smoothed series as TB3_HW. The mean damping parameter is estimated while the trend damping parameter is set to 0.3.

smpl @first @last-10

order.smooth(m,.1,.1,.1) order_hw

smpl @all

graph gra1.line order order_hw

show gra1

smooths the ORDER series by a Holt-Winters multiplicative seasonal method leaving the last 10 observations. The damping parameters are all set to 0.1. The last three lines plot and display the actual and smoothed series over the full sample.

Cross-references