Packet 10 Concept Map

Packet 10 Concept Map

1. Time Series

   • Goal: Generalize the linear regression model to account for dependence of observation
     • ex: consumer price index, temperatures, product consumption
     • Secular trend, cyclical fluctuation, seasonal variation, residual effect
   • Construction of time series
     • 1. E(Yt). 2. Rt 3. Yt = E(Yt) + Rt
     • If evidence of autocorrelation, use a model with...

2. Forecasting

   • Smoothing Techniques
     • Goal: Identify secular trends apart from random fluctuations
     • Moving Averages
     • Exponential Smoothing
   • Accuracy
     • MAPE (mean absolute percentage error)
     • MAD (mean absolute deviation
     • RMSE (root mean squared error)
   • Regression Approach
     • Multilinear regression model with no cyclical component
   • Forecasting with time series autoregressive models
     • Forecasting limits
       • One step
       • Two Step
       • M step

3. Autocorrelation

   • Goal: to correlate time series residuals at different points in time
   • Properties
     • Tends to be positive
     • Diminishes over time
       • First order autoregressive error model
   • Autoregressive error

Images and files in the map