New: Practical Time Series Forecasting with R, 3rd Edition (in full color)
Page/Section
Correction
Chapter 3, Section on Zero Counts (last sentence), p.54
"...positive errors (under-forecasts) than negative errors (over-forecasts)"
Chapter 5, code for creating Figure 5.8
Replace 365.25 with 365 in second line of code to obtain the bottom panel of Figure 5.8 (following a change in the stlm() function)
Chapter 5, pp. 93-95 (equations 5.8, 5.9, 5.13, 5.14)
Replace left-hand-side y(t+1) with y(t)
p. 149, second paragraph
The forecasted residual for April 2001 is computed by plugging in the most recent mean-centered residual from March 2001 (equal to 12.108 - 0.3728) into the AR(1) model: 0.3728 + (0.5998)(12.108 - 0.3728) = 7.411. The positive value tells us that the regression model will produce a ridership forecast for April 2001 that is too low and that we should adjust it up by adding 7,411 riders... the improved two-stage model [regression + AR(1) correction] corrected it by increasing it to 2,011,682 riders.