AdjustTimeSeriesForecast

AdjustTimeSeriesForecast[tproc,forecast,newdata]

adjusts forecast using new observations newdata according to the time series model tproc.

Details

  • AdjustTimeSeriesForecast returns the object of the same type as given in forecast.
  • In AdjustTimeSeriesForecast[tproc,forecast,newdata], forecast can be given in the following forms:
  • {s0,}a path with state si at time i
    {{t0,s0},}a path with state si at time ti
    TemporalData[]one or several paths
  • The times ti and states si must belong to the time and state domain of the process tproc.
  • When the forecast is given as an object containing time stamps, the newdata is aligned according to the time stamps. If the forecast is given as a vector, any time stamps coming with the newdata are ignored, and both the forecast and the newdata are treated as lists of consecutive observations starting at the same point in time. When the newdata carries no time information, the time stamps are created starting with the first time stamp of the forecast.
  • AdjustTimeSeriesForecast may give unreliable results for non-weakly stationary time series models.

Examples

open allclose all

Basic Examples  (3)

Update the forecast for an MA process with one new data point:

Update the forecast for an AR process with two new data points:

Update the forecast for a stationary SARMA process:

Create a forecast for 20 steps ahead:

Update the forecast with two new data points:

Scope  (5)

Input both the forecast and the new data as vectors:

Input the forecast as a vector and new data as TimeSeries:

The time stamps are ignored:

Input the forecast as TemporalData and new data as a vector:

Find the forecast for four steps ahead:

Adjust the forecast with two new data points:

Plot the data, forecast, and adjusted forecast:

Input both the forecast and new data as TemporalData:

Find the forecast for 10 steps ahead:

Define new data starting at time 25:

Find the updated forecast:

The updated forecast remains the same until new data is available:

Adjust a forecast for a vector-valued time series process:

Create a forecast based on data up to time 10:

Use the next two points on the path to update the forecast:

Compare the data with the forecast and the adjusted forecast for each component:

Properties & Relations  (1)

Find the adjusted forecast with mean squared errors:

Find errors of the adjusted forecast:

Possible Issues  (2)

Updating the forecast for a nonstationary time series process may give unreliable results:

Create the forecast for 20 steps ahead:

Adjust the forecast with two new data points:

Mean squared errors are only available when the forecast comes with mean squared errors:

Introduced in 2012
 (9.0)