LOCALTREND Procedure

@LocalTrend performs a "local smoothing regression", which creates an estimate of a trend using fitted values from a locally weighted regression on constant and linear trend.

@LocalTrend(options) x start end y

Parameters

x	input series
start, end	range of x to use. By default, the defined range of x.
y	output series

Options

SPAN=window width [13]

Width of the moving window for the trend regression (should be odd—it will be rounded up to the nearest odd value if it's even). The wider this is, the "stiffer" is the estimated trend function.

Example

* Makridakis et al, Forecasting Methods and Applications, 3rd edition

* Example 3/3 from pp 101-106

open data shampoo.dat

calendar(m) 1990

data(format=free,org=columns) 1990:1 1992:12 shampoo

@localtrend(span=49) shampoo / curve49

@localtrend(span=19) shampoo / curve19

@localtrend(span=7) shampoo / curve7

graph(overlay=dots,ovsame,footer="Very stiff curve") 2

# curve49

# shampoo

graph(overlay=dots,ovsame,footer="Moderately stiff curve") 2

# curve19

# shampoo

graph(overlay=dots,ovsame,footer="Very loose curve") 2

# curve7

# shampoo