UCM Procedure

ESTIMATE Statement

ESTIMATE <options>;

The ESTIMATE statement is an optional statement used to control the overall model-fitting environment. Using this statement, you can control the span of observations used to fit the model by using the SKIPFIRST= and BACK= options. This can be useful in model diagnostics. You can request a variety of goodness-of-fit statistics and other model diagnostic information including different residual diagnostic plots. Note that the ESTIMATE statement is not used to control the nonlinear optimization process itself. That is done using the NLOPTIONS statement, where you can control the number of iterations, choose between the different optimization techniques, and so on. You can save the estimated parameters and other related information in a SAS table by using the OUTEST= option. You can request the optimization of the profile likelihood, the likelihood obtained by concentrating out a disturbance variance, for parameter estimation by using the PROFILE option. The following example illustrates the use of this statement:


estimate skipfirst=12 back=24;

This statement requests that the initial 12 measurements and the last 24 measurements be excluded during the model-fitting process. The actual observation span used to fit the model is decided as follows: Suppose that and are the observation numbers of the first and last nonmissing values of the response variable, respectively. As a result of specifying SKIPFIRST=12 and BACK=24, the measurements between observation numbers and form the estimation span. Of course, the model fitting might not take place if there are insufficient data in the resulting span. The model fitting does not take place if there are regressors in the model that have missing values in the estimation span.

You can specify the following options:

BACK=integer SKIPLAST=integer

indicates that some ending part of the data needs to be ignored during the parameter estimation. This can be useful when you want to study the forecasting performance of a model on the observed data. When BACK=10, the last 10 measurements of the response series are skipped during the parameter estimation. By default, BACK=0.

EXTRADIFFUSE=k

enables continuation of the diffuse filtering iterations for k additional iterations beyond the first instance where the initialization of the diffuse state would have otherwise taken place. If the specified k is larger than the sample size, the diffuse iterations continue until the end of the sample. Note that one-step-ahead residuals are produced only after the diffuse state is initialized. Delaying the initialization leads to a reduction in the number of one-step-ahead residuals available for computing the residual diagnostic measures. This option is useful when you want to ignore the first few one-step-ahead residuals that often have large variance.

LIKE=DIFFUSE | MARGINAL

specifies the type of likelihood to use for parameter estimation. You can specify the following values:

DIFFUSE: uses diffuse likelihood.
MARGINAL: uses marginal likelihood.

By default, LIKE=DIFFUSE. For more information about likelihood types, see the section Likelihood Computation and Model-Fitting Phase in Chapter 14, CSSM Procedure. For an example of the use of LIKE=MARGINAL option, see Example 30.10.

NOPROFILE

requests that the usual likelihood be optimized for parameter estimation. For more information, see the section Parameter Estimation by Profile Likelihood Optimization.

OUTEST=libref.data-table

specifies an output SAS table for the estimated parameters. libref.data-table is a two-level name, where libref refers to the library, and data-table specifies the name of the output data table. For more information about this two-level name, see the section Using CAS Sessions and CAS Engine Librefs.

PLOT=plot-request PLOT=( <plot-request> …<plot-request> )

produces residual diagnostic plots. You can specify the following plot-requests:

ACF

produces a residual-autocorrelation plot.

CUSUM

produces a plot of cumulative residuals against time.

CUSUMSQ

produces a plot of cumulative squared residuals against time.

HISTOGRAM

produces a histogram of residuals.

LOESS

produces a scatter plot of residuals against time, which has an overlaid loess fit.

MODEL

produces a plot of one-step-ahead forecasts in the estimation span.

PACF

produces a residual-partial-autocorrelation plot.

PANEL

produces a summary panel of the residual diagnostics that consists of the following:

histogram of residuals
normal quantile plot of residuals
residual-autocorrelation plot
residual-partial-autocorrelation plot

QQ

produces a normal quantile plot of residuals.

RESIDUAL

produces a needle plot of residuals against time.

WN

produces a plot of p-values, in log scale, at different lags for the Ljung-Box portmanteau white-noise test statistics.

PRINT=NONE

suppresses all the printed output related to the model fitting, such as the parameter estimates, the goodness-of-fit statistics, and so on.

PROFILE

requests that the profile likelihood, obtained by concentrating out one of the disturbance variances from the likelihood, be optimized for parameter estimation. By default, the profile likelihood is not optimized if any of the disturbance variance parameters is held fixed to a nonzero value. For more information see the section Parameter Estimation by Profile Likelihood Optimization.

SKIPFIRST=integer

indicates that some early part of the data needs to be ignored during the parameter estimation. This can be useful if there is a reason to believe that the model being estimated is not appropriate for this portion of the data. When SKIPFIRST=10, the first 10 measurements of the response series are skipped during the parameter estimation. By default, SKIPFIRST=0.

Last updated: November 24, 2025