Examples (GENLINMIXED command)
The following examples correspond to the predefined distribution and link function combinations on the Target settings of the dialog.
Linear model
GENLINMIXED
/FIELDS TARGET=y
/TARGET_OPTIONS DISTRIBUTION=NORMAL LINK=IDENTITY
/FIXED EFFECTS=x1 x2 x3.
- The FIELDS subcommand specifies y as the target.
- The TARGET_OPTIONS subcommand that the target has a normal distribution and is linearly related to the model effects.
- The FIXED subcommand specifies a main effects model with fields x1, x2, and x3. If they are continuous, they will be treated as covariates, if categorical, they will be treated as factors.
Gamma regression
GENLINMIXED
/FIELDS TARGET=claim_amount ANALYSISWEIGHT=number_of_claims
/TARGET_OPTIONS DISTRIBUTION=GAMMA LINK=POWER(-1)
/FIXED EFFECTS=holder_age vehicle_group vehicle_age
/BUILD_OPTIONS INPUTS_CATEGORY_ORDER=DESCENDING.
- The procedure fits a model for the target claim_amount, using holder_age, vehicle_group, and vehicle_age as main effects. In order to account for the varying number of claims used to compute the average claim amounts, you specify number_of_claims as the analysis weight.
- The TARGET_OPTIONS specification assumes that claim_amount has a gamma distribution. A power link function with −1 as the exponent relates the distribution of claim_amount to a linear combination of the predictors, including an intercept term.
- The BUILD_OPTIONS subcommand specifies that the category order for factors is descending values of factor levels; thus, the first category of each categorical field is used as the reference category.
(Poisson) Loglinear model
GENLINMIXED
/FIELDS TARGET=damage_incidents OFFSET=log_months_service
/TARGET_OPTIONS DISTRIBUTION=POISSON LINK=LOG
/FIXED EFFECTS=type construction operation
/BUILD_OPTIONS INPUTS_CATEGORY_ORDER=DESCENDING.
- The procedure fits a model for the target damage_incidents, using type, construction, and operation as main effects.
- The TARGET_OPTIONS specification assumes that damage_incidents has a Poisson distribution. A log link function relates the distribution of damage_incidents to a linear combination of the predictors, including an intercept term, and an offset equal to the values of log_months_service.
- The BUILD_OPTIONS subcommand specifies that the category order for factors is descending values of factor levels; thus, the first category of each categorical field is used as the reference category.
Negative binomial regression
GENLINMIXED
/FIELDS TARGET=damage_incidents OFFSET=log_months_service
/TARGET_OPTIONS DISTRIBUTION=NEGATIVE_BINOMIAL LINK=LOG
/FIXED EFFECTS=type construction operation.
- The negative binomial distribution is an alternative to the Poisson when the observations are overdispersed; that is, since the mean and variance of the Poisson distribution are the same, when the data show greater variability, the negative binomial distribution can provide a better fit.
(Nominal) Multinomial logistic regression
GENLINMIXED
/FIELDS TARGET=bfast
/TARGET_OPTIONS DISTRIBUTION=MULTINOMIAL LINK=LOGIT
/FIXED EFFECTS=agecat gender active.
- The procedure fits a model for bfast using agecat, gender, and active as main effects.
- The TARGET_OPTIONS specification assumes that bfast has a (nominal) multinomial distribution. A logit link function relates the distribution of bfast to a linear combination of the predictors, including an intercept term.
(Ordinal) Multinomial logistic regression
GENLINMIXED
/FIELDS TARGET=chist
/TARGET_OPTIONS DISTRIBUTION=MULTINOMIAL LINK=CLOGLOG
/FIXED EFFECTS=numcred othnstal housng age duration.
- The procedure fits a model for chist using numcred, othnstal, housing, age, and duration as main effects. Because numcred, othnstal, and housing have categorical measurement level, they are treated as factors; age, and duration have continuous (scale) measurement level and are treated as covariates.
- The TARGET_OPTIONS specification assumes that chist has an (ordinal) multinomial distribution. A (cumulative) complementary log−log link function relates the distribution of chist to a linear combination of the predictors, including threshold terms for the categories of chist (except the last category).
Binary logistic regression
GENLINMIXED
/FIELDS TARGET=default
/TARGET_OPTIONS DISTRIBUTION=BINOMIAL LINK=LOGIT
/FIXED EFFECTS=age ed employ address income debtinc creddebt othdebt.
- The procedure fits a model for default using age, ed, employ, address, income, debtinc, creddebt, and othdebt as main effects.
- The TARGET_OPTIONS specification assumes that default has a binomial distribution. A logit link function relates the distribution of default to a linear combination of the predictors, including an intercept term.
Binary probit model
GENLINMIXED
/FIELDS TARGET=response TRIALS=nsubj
/TARGET_OPTIONS DISTRIBUTION=BINOMIAL LINK=PROBIT
/FIXED EFFECTS=site value.
- The procedure fits a model for the number of responses response within trials nsubj using site and value as main effects.
- The TARGET_OPTIONS specification assumes that response within nsubj has a binomial distribution. A probit link function relates the distribution of the target to a linear combination of the predictors, including an intercept term.
Interval censored survival
GENLINMIXED
/FIELDS TARGET=result2
/TARGET_OPTIONS DISTRIBUTION=BINOMIAL LINK=CLOGLOG
/FIXED EFFECTS=duration treatment period age USE_INTERCEPT=FALSE
/BUILD_OPTIONS TARGET_CATEGORY_ORDER=DESCENDING INPUTS_CATEGORY_ORDER=DESCENDING.
- The procedure fits a model for the target result2, using duration, treatment, period, and age as main effects.
- The BUILD_OPTIONS subcommand specifies that the category order for the target and all factors is descending values of factor levels; thus, the first category of each categorical field is used as the reference category.
- The TARGET_OPTIONS specification assumes that result2 has a binomial distribution. A complementary log-log link function relates the probability of result2 to a linear combination of the predictors, excluding an intercept term.
Linear mixed model
GENLINMIXED
/DATA_STRUCTURE SUBJECTS=school*classroom*student_id
/FIELDS TARGET=posttest TRIALS=NONE OFFSET=NONE
/TARGET_OPTIONS DISTRIBUTION=NORMAL LINK=IDENTITY
/FIXED EFFECTS=school_setting school_type teaching_method n_student
gender lunch pretest
USE_INTERCEPT=TRUE
/RANDOM USE_INTERCEPT=TRUE SUBJECTS=school
COVARIANCE_TYPE=VARIANCE_COMPONENTS
/RANDOM USE_INTERCEPT=TRUE SUBJECTS=school*classroom
COVARIANCE_TYPE=VARIANCE_COMPONENTS
/BUILD_OPTIONS TARGET_CATEGORY_ORDER=ASCENDING
INPUTS_CATEGORY_ORDER=ASCENDING MAX_ITERATIONS=100
CONFIDENCE_LEVEL=95 DF_METHOD=RESIDUAL COVB=MODEL
/EMMEANS_OPTIONS SCALE=ORIGINAL PADJUST=LSD.
- The DATA_STRUCTURE subcommand specifies that subjects are defined by school, classroom, and student_id.
- The FIELDS and TARGET_OPTIONS subcommands specify that the procedure creates a model for posttest using a normal distribution to fit the test score and an identity link to relate the target to a linear combination of the predictors.
- The FIXED subcommand speficies a model with school_setting, school_type, teaching_method, n_student, gender, lunch, and pretest as main effects.
- The first RANDOM subcommand specifies an intercept-only random effect block with school as the subject field. This should account for correlation between classrooms within the same school.
- The second RANDOM subcommand specifies an intercept-only random effect block with school*classroom as the subject field. This should account for correlation between students within the same classroom.
- All other options are set to their default values.
For discussion of output from this syntax and to place the example in context, see Analyzing test scores .
Poisson loglinear mixed model
GENLINMIXED
/DATA_STRUCTURE SUBJECTS=patient_id
/FIELDS TARGET=convulsions TRIALS=NONE OFFSET=NONE
/TARGET_OPTIONS DISTRIBUTION=POISSON LINK=LOG
/FIXED EFFECTS=after_t treatment*after_t USE_INTERCEPT=TRUE
/RANDOM EFFECTS=after_t USE_INTERCEPT=TRUE
SUBJECTS=patient_id COVARIANCE_TYPE=UNSTRUCTURED
/BUILD_OPTIONS TARGET_CATEGORY_ORDER=ASCENDING
INPUTS_CATEGORY_ORDER=ASCENDING MAX_ITERATIONS=100
CONFIDENCE_LEVEL=95 DF_METHOD=RESIDUAL COVB=MODEL
/EMMEANS_OPTIONS SCALE=ORIGINAL PADJUST=LSD.
- The DATA_STRUCTURE subcommand specifies that subjects are defined by patient_id.
- The FIELDS and TARGET_OPTIONS subcommands specify that the procedure creates a model for convulsions using a Poisson distribution to fit the number of convulsions and a log link to relate the target to a linear combination of the predictors.
- The FIXED subcommand specifies a model with after_t and treatment*after_t as effects.
- The RANDOM subcommand specifies after_t and an intercept as effects in a random effect block with patient_id as the subject field. This should account for correlation between repeated observations of the same patient.
- All other options are set to their default values.
For discussion of output from this syntax and to place the example in context, see Determining treatment effectiveness in a clinical trial .
Multinomial logistic mixed model
GENLINMIXED
/DATA_STRUCTURE SUBJECTS=customer_id
/FIELDS TARGET=service_usage TRIALS=NONE OFFSET=NONE
/TARGET_OPTIONS DISTRIBUTION=MULTINOMIAL LINK=LOGIT
/FIXED EFFECTS=edcat inccat reside service_type
USE_INTERCEPT=TRUE
/RANDOM USE_INTERCEPT=TRUE SUBJECTS=customer_id
COVARIANCE_TYPE=VARIANCE_COMPONENTS
/BUILD_OPTIONS TARGET_CATEGORY_ORDER=ASCENDING
INPUTS_CATEGORY_ORDER=ASCENDING MAX_ITERATIONS=100
CONFIDENCE_LEVEL=95 DF_METHOD=RESIDUAL COVB=MODEL
/EMMEANS_OPTIONS SCALE=ORIGINAL PADJUST=LSD.
- The DATA_STRUCTURE subcommand specifies that subjects are defined by customer_id.
- The FIELDS and TARGET_OPTIONS subcommands specify that the procedure creates a model for service_usage using a multinomial distribution and a logit link to relate the target to a linear combination of the predictors.
- The FIXED subcommand specifies a model with edcat, inccat, reside, and service_type as main effects.
- The RANDOM subcommand specifies an intercept-only random effect block with customer_id as the subject field. This should account for correlation between answers to the service usage questions across service types (tv, phone, internet) within a given survey responder's answers.
- All other options are set to their default values.
For discussion of output from this syntax and to place the example in context, see Profiling cable customers .