Single Series ARIMA Results - Advanced Tab

Select the Advanced tab of the Single Series ARIMA Results dialog box to access the options described here.

Element Name Description
Summary: Parameter estimates Click the Summary: Parameter estimates button to display a spreadsheet with the parameter estimates, their standard errors, the t-values (parameter estimates divided by standard errors), and the respective p-values. Note that the standard errors are approximations, computed from the inverse of the matrix of partial derivatives approximated via finite differencing (refer to the Overview). When an interrupted time series is analyzed, this spreadsheet will also contain the parameter estimates for Delta and Omega. For more details, see impact patterns.
Print results Click the Print results button to display a summary of the results in a report. To send the results to a printer, see Printing a Report.
Parameter covariances/correlations Click the Parameter covariances/correlations button to display a spreadsheet containing parameter covariances and a spreadsheet containing parameter correlations. The variance/covariance matrix of parameter estimates is computed from a finite difference approximation of the Hessian matrix of partial derivatives. The resulting estimates are asymptotic in nature. The higher the correlation between two parameters, the greater is their redundancy of their contribution to the fit of the model. The parameter covariance matrix cannot be computed if the Hessian matrix cannot be inverted, which usually indicates that one or more parameters are completely redundant. In terms of the ARIMA model, this usually means that the model is inappropriate (misspecified). In that case, carefully review the patterns of (partial) autocorrelation and try to fit a simpler model.
Forecasting Use the options in the Forecasting group box to compute ARIMA forecasts, which are computed after all transformations and differencing (selected on the Single Series ARIMA dialog box) have been "undone"; that is, the computed forecasts will be expressed in the units of the original time series. The computations of the forecasts and their standard errors closely follows the algorithm described in Box & Jenkins (1976; Program 4).
Forecast cases Click the Forecast cases button to display a spreadsheet with 1) the ARIMA forecasts, 2) their standard errors (if no log or power transformation was requested), 3) their confidence limits according to the value in the Confidence level box (see below), and 4) the observed values and residuals (observed minus predicted), if applicable (i.e., if forecasts are requested for cases that were also observed; see below).
Plot series & forecasts Click the Plot series & forecasts button to display a line graph showing 1) the values of the observed series, 2) the values of the forecasts, and 3) the confidence limits around the forecasts, according to the value in the Confidence level box (see below). The scaling and labeling of the horizontal (X) axis in this plot depend on the settings in the Label data points with group box on the Review & residuals tab.
Number of cases; Start at case Enter values in the Number of cases box and Start at case box to determine 1) how many forecasts will be computed, and 2) at which point in the series the forecasts begin. Ordinarily, you want to predict the future, so by default STATISTICA will compute the forecasts starting with the first observation in the future (after the last valid observed case) through at least one seasonal cycle. However, a very powerful visual way of evaluating the quality of the fit of the current ARIMA model to the data is to evaluate how well the predicted values follow the actual observed values. Therefore, it is sometimes useful to compute forecasts for two or three seasonal cycles in the middle of the series.
Note: the Start at case parameter should always be set in terms of the original case numbers in the file, not in terms of valid case numbers for the respective series. For example, if the variable being analyzed has, in the data file, missing data for the first 20 cases, and then 100 valid cases, then the first future case will be 20+100+1=121 (even though there are only 80 valid cases in the series).
Confidence level Enter a value in the Confidence level box to determine the confidence band around the forecasts in the spreadsheets and plots.
Append forecasts to original series on Exit If you select the Append forecasts to original series check box (or leave it at its default setting), then when exiting this dialog the forecasts (the ones computed the last time prior to Canceling) will be combined with the observed values (up to the forecasts), and this new series will be appended to the active work area. So for example, if you requested 20 forecast starting with case number 101, the first 101 observations for the new series will be copied from the original series, and then 20 forecasts (case 101 to 120) will be added.