Chapter 12 Time Series Analysis and Forecasting
Extrapolation methods
Forecasting methods where only past values of a variable (and possible time itself) are used to forecast future values
Seasonality
A regular pattern of ups and downs based on the season of the year, typically months or quarters
Smoothing constants
Constants between 0 and 1 that prescribe the weight attached to previous observations and hence the smoothness of the series of forecasting
Correlogram
A bar chart of autocorrelations at different lags
Exponential smoothing models
A class of forecasting models where forecasts are based on weighted averages of previous observations, giving more weight to more recent observations
Moving averages model
A forecasting model where the average of several past observations is used to forecast the next observation
Ratio-to-moving averages method
A method for deseasonalizing a time series, so that some other method can then be used to forecast the deseasonalized series
Random walk model
A model indicating that the differences between adjacent observations of a time series variable are constant except for random noise
Linear trend model
A regression model where a time series variable changes by a constant amount each time
Exponential trend model
A regression model where a time series variable changes by a constant percentage each time period
Autoregression model
A regression model where the only explanatory variables are lagged values of the dependent variable (and possibly other time series variables or their lags)
Dummy variables for seasonality
A regression-based method for forecasting seasonality, where dummy variables are used for the season
Trend
A systematic increase or decrease of a time series variable through time
Runs test
A test of whether the forecast errors are random noise
Simple exponential smoothing
An exponential smoothing model useful for time series with no prominent trend or seasonality
Cyclic variation
An irregular pattern of ups and downs caused by business cycles
Autocorrelations
Correlations of a time series variable with lagged versions of itself
Causal (or econometric) methods
Forecasting methods based on regression, where other time series variables are used as explanatory variables
Mean absolute error (MAE)
The average of the absolute forecast errors
Mean absolute percentage error (MAPE)
The average of the absolute percentage forecast errors
Forecast error
The difference between the actual value and the forecast
Span
The number of observations in each average of a moving averages model
Root mean square error (RMSE)
The square root of the average of the squared forecast errors
Noise (or random variation)
The unpredictable ups and downs of a time series variable