Overview (GENLINMIXED command)
The GENLINMIXED
procedure fits generalized linear mixed models. Generalized linear
mixed models extend the linear model so that:
- The target is linearly related to the factors and covariates via a specified link function
- The target can have a non-normal distribution
- The observations can be correlated.
Generalized linear mixed models cover a wide variety of models, from simple linear regression to complex multilevel models for non-normal longitudinal data.
Options
- Probability Distribution of Target
- The probability distribution of the target can be specified as normal, binomial, gamma, inverse Gaussian, multinomial, negative binomial, or Poisson.
- Events/Trials Target Specification
- The typical target specification will be a single field, but for the binomial distribution the target can be specified using a number-of-events field and a number-of-trials field. Alternatively, if the number of trials is the same across all subjects, then trials can be specified using a fixed number instead of a field.
- Link Function
- The target is linearly related to the model effects via one of the following link functions: Identity, complementary log-log, log, log-complement, logit, negative log-log, power, and probit.
- Repeated Measurements
- You can specify the subject structure for repeated measurements and how the errors of the repeated measurements are correlated. Complete independence is assumed across subject blocks.
- Knonecker Measurements
- You can specify the subject structure for Knonecker measurements and how the
errors of the measurements are correlated. Knonecker should only be used when
COVARIANCE_TYPE
is defined asUN_AR1
,UN_CS
, orUN_UN
. - Random Effects
- Random-effects are effects whose values in the data file can be considered a random sample from a larger population of values. They are useful for explaining excess variability in the target.
- Estimated Marginal Means
- Estimated marginal means can be computed for one or more crossed factors and can be based on either the response or the linear predictor.
- Output
- The procedure produces a model viewer object containing tables and charts for assessing the quality of the model; also you can save predicted values to the active dataset and the model to a file in PMML format.
Basic specification
The basic specification is the GENLINMIXED
command and the
FIELDS
subcommand with the TARGET
keyword. This fits an
intercept-only model that can be used as a basis for comparing more complicated models.
Note: Measurement level can affect the results. If any variables (fields) have an unknown
measurement level, a data pass is performed to determine the measurement level before the analysis
begins. For information on the determination criteria, see SET SCALEMIN.
Syntax rules
- The
FIELDS
subcommand is required. All other subcommands are optional. - Subcommands can be specified in any order.
- There can be multiple
RANDOM
andEMMEANS
subcommands; eachRANDOM
subcommand defines a separate random effects block, while eachEMMEANS
subcommand defines a separate set of estimated means. Only a single instance of any other subcommand is allowed. - An error occurs if a keyword is specified more than once within a subcommand.
- Parentheses, equals signs, and slashes shown in the syntax chart are required.
- The command name, subcommand names, and keywords must be spelled in full.
- Empty subcommands are not allowed.
- Any split field defined on the
SPLIT FILE
command cannot be used on this command. - Any field specifed on the
WEIGHT
command cannot be used on this command.
Limitations
- The
SPLIT FILE
setting is ignored with a warning by the procedure.