The SSM Procedure
Details
- State Space Model and Notation
- Types of Sequence Data
- Overview of Model Specification Syntax
- Filtering, Smoothing, Likelihood, and Structural Break Detection
- Estimation of User-Specified Linear Combination of State Elements
- Contrasting PROC SSM with Other SAS Procedures
- Predefined Trend Models
- Predefined Structural Models
- Models with Dependent Lags
- Temporal Aggregation and Temporal Distribution
- Covariance Parameterization
- Missing Values
- Computational Issues
- Displayed Output
- ODS Table Names
- ODS Graph Names
- OUT= Data Set
Throughout this section, vectors and matrices are denoted by boldface letters. Generally, Greek letters (such as ,
, and
) denote unobserved or latent quantities—often estimated from the data—that represent model parameters, latent states, or noise variables. Capital letters such as X, Y, and Z are used to denote the observed data variables. Whenever there is no ambiguity, it is assumed that the matrices have appropriate dimensions when they are being multiplied—in particular, the vectors behave as column vectors or row vectors as the need arises. On many occasions, matrices are described inline—that is, they are described as parenthesized lists, in a rowwise fashion, with the rows separated by a comma. The term "dot product" is used to describe the scalar that results from the product of a row vector with a (conforming) column vector.