contents.gifindex.gifprev1.gifnext1.gif

Regression Trend Lines

Some of the forecast types you can choose are presented below. All of these forecasts are fit to your data using regression analysis.

Linear

dph00217.gif

dph00218.gifdph00219.gif

Linear forecast with confidence limits

dph00220.gif

n is a user-supplied constant between 0 and 100 defining the confidence interval. For example, a value of 90 will cause confidence lines to be drawn such that the value being forecast is 90% likely to be between the lines (5% chance of being above the upper line and 5% below the lower). To display lines at one standard deviation from the prediction line, use n = 68.3. To display lines at two standard deviations, use n = 95.5%.

In order for the confidence limits to properly reflect uncertainty in the forecast, the underlying data must conform to assumptions inherent in the linear regression model. Specifically, y must be a linear function of x; and, the residuals must be normally distributed.

dph00221.gifdph00222.gif

Hyperbolic

dph00223.gif

dph00224.gifdph00225.gif

Logarithmic

dph00226.gif

dph00227.gifdph00228.gif

Square Root

dph00229.gif

dph00230.gifdph00231.gif

Quadratic

dph00232.gif

dph00233.gifdph00234.gif

Polynomial

dph00235.gif  n is a user-supplied constant.

dph00236.gifdph00237.gif

Power

dph00238.gif

dph00239.gifdph00240.gif

Exponential

dph00241.gif

dph00242.gifdph00243.gif

Seasonal trend

dph00244.gif

p1 and p2 are user-supplied season and cycle periods. All other constants are calculated automatically.

dph00245.gif