GENLINMIXED

GENLINMIXED is available in the Advanced Models option.

The GENLINMIXED procedure fits generalized linear mixed models.

GENLINMIXED
  [/DATA_STRUCTURE]
     SUBJECTS = fieldName[*fieldName[*...]]
     [REPEATED_MEASURES = fieldName[*fieldName[*...]]]
     [KRONECKER_MEASURES = fieldName[*fieldName[*...]]]
     [GROUPING = fieldName[*fieldName[*...]]]
     [COVARIANCE_TYPE = DIAGONAL** | AR1 | ARH1 |ARMA11 | COMPOUND_SYMMETRY | CSH | IDENTITY |
                        SP_POWER | SP_EXPONENTIAL | SP_GAUSSIAN | SP_LINEAR | SP_LINEARLOG |
                        SP_SPHERICAL | TOEPLITZ | UN_AR1 | UN_CS | UN_UN | UNSTRUCTURED | 
                        VARIANCE_COMPONENTS]
     [SPCOORDS = fieldName...]
  /FIELDS
    TARGET = fieldName
    [TRIALS = {NONE**          }]
              {VALUE(number)   }
              {FIELD(fieldName)}
    [ANALYSIS_WEIGHT = fieldName]
    [OFFSET = {NONE**          }]
              {VALUE(number)   }
              {FIELD(fieldName)}
  [/TARGET_OPTIONS]
    [REFERENCE = value]
    [DISTRIBUTION = NORMAL | BINOMIAL | MULTINOMIAL | GAMMA | 
                    INVERSE_GAUSSIAN | NEGATIVE_BINOMIAL | POISSON]
    [LINK = IDENTITY | CAUCHIT | CLOGLOG | LOG | LOGC | LOGIT | NLOGLOG | POWER | PROBIT ]
    [LINK_PARAMETER = number]
  [/FIXED]
    [EFFECTS = effect [effect ...]]
    [USE_INTERCEPT = TRUE** | FALSE]
  [/RANDOM]
    [EFFECTS = effect [effect ...]]
    [USE_INTERCEPT = TRUE | FALSE**]
    [SUBJECTS = fieldName[*fieldName[*...]]]
    [GROUPING = fieldName[*fieldName[*...]]]
    [COVARIANCE_TYPE = DIAGONAL | AR1 | ARH1 | ARMA11 | COMPOUND_SYMMETRY | CSH |
                       IDENTITY | TOEPLITZ | UNSTRUCTURED | VARIANCE_COMPONENTS**]
    [SOLUTION = TRUE | FALSE**]
  [/RANDOM ...]
  [/BUILD_OPTIONS]
    [TARGET_CATEGORY_ORDER = {ASCENDING**}]
                             {DESCENDING }
                             {DATA       }
    [INPUTS_CATEGORY_ORDER = {ASCENDING**}]
                             {DESCENDING }
                             {DATA       }
    [MAX_ITERATIONS = {100** }]
                      {number}
    [CONFIDENCE_LEVEL = {95**  }]
                        {number}
    [DF_METHOD = {RESIDUAL**   }]
                 {SATTERTHWAITE}
                 {KENWARD_ROGER}
    [COVB = {MODEL**}]
            {ROBUST }
    [PCONVERGE = {0**   }[({ABSOLUTE**})] ] 
                 {value }  {RELATIVE  }
    [LCONVERGE = {0**  }[({ABSOLUTE**})] ]
                 {value}  {RELATIVE  }
    [HCONVERGE = {.00000001**}[({ABSOLUTE  })] ]
                 {value      }  {RELATIVE**}
    [SCORING = {0**}]
               {n  } 
    [SINGULAR = {1E-12**}]
                {value  }
  [/EMMEANS 
    [TABLES = factor[*factor...]]
    [COMPARE = factor]
    [CONTRAST = {NONE**   }]
                {PAIRWISE }
                {DEVIATION}
                {SIMPLE   }
  [/EMMEANS...]
  [/EMMEANS_OPTIONS]
    [CONTROL = fieldName (number) [fieldName (number) ...]]
    [SCALE = {ORIGINAL** }]
             {TRANSFORMED}
    [PADJUST = {LSD**        }]
               {SEQBONFERRONI}
               {SEQSIDAK     }
  [/[/OUTFILE [MODEL = 'file']]
    [EBLUPS = 'savfile' | dataset]
    [FILE_SEPARATE = {FALSE**} {TRUE}]]
  [/SAVE [PREDICTED_VALUES(fieldName)]]
    [PREDICTED_PROBABILITY(rootname)]
    [MAX_CATEGORIES({25**   })]
                    {integer}
    [CONFIDENCE(fieldName)]
    [CONFIDENCE_METHOD({PREDICTEDPROBABILITY**})]
                       {PROBABILITYIMPROVEMENT}
    [CONFIDENCE_INTERVALS(rootname)]
    [PEARSON_RESIDUALS(fieldName)].

** Default if the subcommand or keyword is omitted.

The KRONECKER_MEASURES keyword should be only when COVARIANCE_TYPE is defined as UN_AR1, UN_CS, or UN_UN.

When both KRONECKER_MEASURES and REPEATED_MEASURES keywords are in effect, they may or may not have common fields, but their values cannot be exactly the same (even when the values are not in the same order).

This command reads the active dataset and causes execution of any pending commands. See the topic Command Order for more information.

Syntax for the GENLINMIXED command can be generated from the Generalized linear mixed models dialog.

Release History

Release 19

  • Command introduced.

Release 20

  • Ordinal targets can be analyzed using the Multinomial distribution and the complementary log-log, cauchit, logit, negative log-log, or probit link functions.

Release 22

  • PCONVERGE, LCONVERGE, HCONVERGE, SCORING, and SINGULAR keywords introduced on the BUILD_OPTIONS subcommand.

Release 25

  • SP_POWER, SP_EXPONENTIAL, SP_GAUSSIAN, SP_LINEAR, SP_LINEARLOG, and SP_SPHERICAL options added to the COVARIANCE_TYPE keyword on the DATA_STRUCTURE subcommand.
  • SPCOORDS keyword introduced on the DATA_STRUCTURE subcommand.
  • SOLUTION keyword introduced on the RANDOM subcommand.

Release 28.0.1

  • Default convergence criterion in effect changed from PCONVERGE=.000001 (ABSOLUTE) to HCONVERGE=.00000001 (RELATIVE).

Subscription (April 2018 update)

  • ARH1 and CSH options added to the COVARIANCE_TYPE keyword on the RANDOM subcommand.
  • ARH1 and CSH options added to the COVARIANCE_TYPE keyword on the DATA_STRUCTURE subcommand.
  • KENWARD_ROGER option added to the DF_METHOD keyword on the BUILD_OPTIONS subcommand.
  • KRONECKER_MEASURES keyword added to the DATA_STRUCTURE subcommand. The keyword should be used only when COVARIANCE_TYPE is one of three following Kronecker types.
  • UN_AR1, UN_CS and UN_UN options added to the COVARIANCE_TYPE keyword on the DATA_STRUCTURE subcommand.

Release 29.0.0

  • The output includes pseudo-R2 measures and the intra-class correlation coefficient (when appropriate).

Release 29.0.1

  • EBLUPS and FILE_SEPARATE keywords added to the OUTFILE subcommand.

Example

GENLINMIXED
  /FIELDS TARGET=y1.