Simple Regression 单的回归.docVIP

Simple Regression (Trend line Analysis) Ft = a + b t Ft = forecast for time t a = y intercept of the line b = the slope of the line n = # of observations Single Moving Averages Long Form: N = # periods in moving average Short Form: Weighted Moving Average Average Percentage Change Moving Average Percentage Change Moving Average Percentage Change Single Exponential Smoothing Smoothed Percentage Change Adaptive Response Rate Single Exponential Smoothing (ARRSES) 0 ≤ ?≤ 1 Linear Moving Average (Double Moving Average) Brown’s One Parameter Linear Exponential Smoothing (Double ES) Holt’s Two – Parameter Linear Exponential Smoothing (Formula for initial conditions- copy here) Brown’s Quadratic Exponential Smoothing (Triple Smoothing) Winter’s Exponential Smoothing Relative Parameter Values for Moving Average and Exponential Smoothing ? Low High Low Medium # of MA terms Medium α value High # of MA terms Low α value High Low # of MA terms High α value Medium # of MA terms Medium α value Classical Decomposition Method Ratio–to–Moving Averages 1. 2. 3. Calculate a trend value Divide the moving average by the trend value eliminates trend isolates cycle To Prepare a forecast: A trend value is calculated for period to be forecast.  Tt = a + b t Multiply this by the appropriate seasonal factor It Multiply this by the appropriate cyclical factor (may not be easy because length of cycle may vary) Ct Ft = Tt x It x Ct SPCt = SPCt-1 + ( (PCt – SPCt-1) Ft+m = (SPCt x (m) x (Xt) ) + Xt = ( (SPCt x m) + 1 ) Xt Ft+1 = Ft + ( (Xt - Ft) 0 ≤ ( ≤ 1 Assume: F1 = X1 Can use Short Form to get MPCt Short Form: Long Form: m = # of periods ah