MODEL Subcommand (SURVREG RECURRENT command)

MODEL is a required subcommand to specify the model options and settings. It might optionally be followed by model effects and initial values. If no model effects are specified, all the variables that are specified after SURVREG RECURRENT BY and WITH is used as main effects. If specified, the variable(s) must be contained in the factors that are specified after SURVREG RECURRENT BY or the covariates that are specified after SURVREG RECURRENT WITH. A factor effect might optionally be followed by multiple numeric values that are specified within brackets. If specified, the adjacent values might be separated by a comma, a blank space, or blank spaces. Each one is then used as the initial value for the estimation of the corresponding coefficient of the category. A default value, denoted by DEFAULT, might be specified at a given position to and assigned to a certain category. All the values that are specified within brackets cannot be DEFAULT. If the number of values (including the default values) does not match the one of the categories (including the reference category), the procedure then ignores all the values that are specified for the factor and assign default values to all the categories. A covariate effect might optionally be followed by a single numeric value that is specified within brackets. If specified, it is used as the initial value for the estimation of the corresponding coefficient. Specifying DEFAULT within brackets is not allowed for a covariate.

SUBJECT

A required parameter to specify the subject ID. It must be a single variable.

INTERVAL

An optional parameter to specify the interval number used to identify the different recurrent records that share subject ID. If specified, it must be a single and numeric variable and the order of a subject’s records is designated by it. With INTERVAL in effect, for a given subject, all the records that follow the first failure would be excluded from the analysis. If INTERVAL is not specified, all the records are retained.

DISTRIBUTION

An optional parameter to specify the distribution of the survival time. WEIBULL is the default setting, which specifies Weibull distribution. EXPONENTIAL specifies exponential distribution. LOG_NORMAL specifies log-normal distribution. LOG_LOGISTIC specifies log-logistic distribution.

SIGMA

An optional parameter to control the setting of the scale parameter. SE is the default setting, which uses the standard error of the corresponding ordinary least squares regression as the initial value. INV_SE uses the reciprocal of the standard error. If a single numeric value is specified, the value is used as the initial value. If specified, it must be greater than 0. The process is ignored if EXPONENTIAL is specified for DISTRIBUTION.

INTERCEPT

An optional parameter to specify the initial value of the intercept term. If specified, it must be a single numeric value, and cannot be 0.

FRAILTY

An optional parameter to specify the distribution of the frailty component. GAMMA is the setting by default, which specifies the Gamma distribution. INV_GAUSSIAN specifies the inverse-Gaussian distribution.

VAR_FRAILTY

An optional parameter to specify the initial value of the variance of the frailty component. It must be a single numeric value greater than 0. The default value is 1.0 if FRAILTY=GAMMA and 0.1 if FRAILTY=INV GAUSSIAN.