This preview has intentionally blurred sections. Sign up to view the full version.
Unformatted text preview: time horizon t.window → what window size you want, 15 day moving average would be t.window = 15 Doing the moving average of the moving average lets you do an even number of moving avgs To figure out which moving average is best, compute the sum of the square of errors summation(Ei^2) Ei = Yi - estimated Yi Ways of forecasting We have y1, y2, y3, and the forecast Naive → forecast = y3 (the previous one is the only one that matters) average → forecast = (y1+y2+y3)/3 (all weighted the same) exponential → forecast = w1y1 + w2y2 + w3y3 (all weighted differently) Gets the weight by figuring out the least sum of square errors When forecasting, figuring out the trend and the seasonal component, not the noise...
View Full Document
- Fall '13
- Regression Analysis, residual sum, squared errors, seasonal variation, multiplicative decomposition, average → forecast