Publication date: 10/01/2019

Model Reports

The Model Reports report contains a Select Series list and one or more individual series modeling reports. The individual reports that are displayed are determined by the Select Series list. Click on the name of a series in this list to show the report for the selected series. If you select more than one series name by using either Ctrl or Shift, an individual model fit report for each of the selected series is shown. If you deselect one or more series from the Select Series list, the reports for the corresponding series are removed.

Each individual model report contains the following information:

Model Type

Describes the best fitting model for the series in terms of error, trend, and seasonality.

Fitted Model Using Entire Series

A group of reports based on the model that is fit to the entire series.

Original Series and Forecasts

A plot of the original series and forecasts. During the time period of the original series, the one-step-ahead predictions are shown. Beyond the time period of the original series, the forecasts are shown. The shaded region represents the forecasting interval.

Model Summary

Provides the -2Loglikelihood, the AIC, the number of parameters, and the sigma value. The sigma value is the estimate of the one-step-ahead forecast standard deviation.

Parameter Estimates

Provides the parameter estimates for the model. The parameters correspond to the parameters in the best fitting model, which is described in the Model Type report.

One-Step-Ahead Forecasting Errors

A plot of the one-step-ahead forecasting errors over time. For each time point t, the error is calculated as e(t) = y(t) - μ(t).

One-Step-Ahead Relative Forecasting Errors for Multiplicative Error Model

(Available only for multiplicative error models.) A plot of the one-step-ahead relative forecasting errors over time. For each time point t, the error is calculated as e(t) = [y(t) - μ(t)]/μ(t).

Fitted Model with Training - Holdout Split

(Appears only when a nonzero value is specified for NHoldout.) A group of reports based on the model that is fit to the training series.

Training and Holdout Series

A plot of the training series and the holdout series. During the time period of the training series, the one-step-ahead predictions are shown. During the time period of the holdout series, the forecasts are shown. The shaded region represents the forecasting interval.

Parameter Estimates using Training Series

Provides the parameter estimates for the model. The parameters correspond to the parameters in the best fitting model, which is described in the Model Type report.

One-Step-Ahead Forecasting Errors

A plot of the one-step-ahead forecasting errors over time, for the training series only. For each time point t, the error is calculated as e(t) = y(t) - μ(t).

One-Step-Ahead Relative Forecasting Errors for Multiplicative Error Model

(Available only for multiplicative error models.) A plot of the one-step-ahead relative forecasting errors over time, for the training series only. For each time point t, the error is calculated as e(t) = [y(t) - μ(t)]/μ(t).

Forecasting Errors on Holdout Series

A plot of the difference between the actual values of the holdout series and the forecasted values from the model.

Want more information? Have questions? Get answers in the JMP User Community.