Decision Analysis

Time series vs casual method

Time series = one variable over time; causal = multiple variables explaining why.

trend vs seasonality

Trend = overall direction long term. Seasonality = repeating pattern short term

Forecast accuracy methods (forecast error, absolute error, squared error, absolute percentage error)

Forecast Error → direction of error, not a very useful measure
Absolute Error → absolute value of forecast errors
Squared Error → squared forecast errors
APE → absolute percentage errors of the forecasts, depends on scale of the data

Averages (simple moving average, weighted moving average, exponential smoothing)

SMA → equal weights
WMA → custom weights
Exponential → select only the weight for the most recent observation

objective of time series analysis

to identify patterns in past data so you can make accurate future forecasts

curve fitting method

minimizes the sum of squared errors between actual data and predicted values.

time series plot

geographical presentation of relationship between time and time series

linear trend

straight line increase/decrease

power trend

curved line (growth changes overtime)

exponential trend

rapid growth or decay

deseasonalizing time series

removing seasonal effects to better see the true trend and underlying pattern

seasonal index

1=above average, less than 1=below average