Variable List (GENLIN command)

The GENLIN command variable list specifies the dependent variable using either a single variable or events and trials variables. Alternatively, the number of trials may be specified as a fixed number. The variable list also specifies any factors and covariates.

If an events/trials specification is used for the dependent variable, then the GENLIN procedure automatically computes the ratio of the events variable over the trials variable or number. Technically, the procedure treats the events variable as the dependent variable in the sense that predicted values and residuals are based on the events variable rather than the events/trials ratio.

  • The first specification on GENLIN must be a single dependent variable name or an events/trials specification.
  • If the dependent variable is specified as a single variable, then it may be scale, an integer-valued count variable, binary, or ordinal.
  • If the dependent variable is binary, then it may be numeric or string and there may be only two distinct valid data values.
  • If the dependent variable is categorical, then it may be numeric or string and must have at least two distinct valid data values.
  • If the dependent variable is not binary or categorical, then it must be numeric.

    REFERENCE Keyword

    The REFERENCE keyword specifies the dependent variable value to use as the reference category for parameter estimation. No model parameters are assigned to the reference category.

    LAST. The last dependent variable value is the reference category. The last dependent variable value is defined based on the ascending order of the data values. This is the default. If REFERENCE = LAST, then the procedure models the first value as the response, treating the last value as the reference category.

    FIRST. The first dependent variable value is the reference category. The first dependent variable value is defined based on the ascending order of the data values. If REFERENCE = FIRST, then the procedure models the last value as the response, treating the first value as the reference category.

    value. The specified dependent variable value is the reference category. Put the value inside a pair of quotes if it is formatted (such as date or time) or if the dependent variable is of string type; note, however, that this does not work for custom currency formats. If REFERENCE = value, then the procedure models the unspecified value as the response, treating the specified value as the reference category.

    • The REFERENCE specification is honored only if the dependent variable is binary and the binomial distribution is used (that is, DISTRIBUTION = BINOMIAL is specified on the MODEL subcommand). Otherwise, this specification is silently ignored.
    • If the dependent variable is a string variable, then the value at the highest or lowest level is locale-dependent.
    • If a value is specified as the reference category of the dependent variable, then the value must exist in the data.

    ORDER Keyword

    The ORDER keyword following the dependent variable is honored only if the dependent variable is categorical and the multinomial distribution is used (/MODEL DISTRIBUTION = MULTINOMIAL). Otherwise, this specification is silently ignored. ORDER determines the sort order of the dependent variable’s values. Cumulative link functions are applied based on this order.

    ASCENDING. Dependent variable values are sorted in ascending order, from the lowest value to the highest value. This is the default.

    DATA. Dependent variable values are not sorted. The first value encountered in the data defines the first category, the last value encountered defines the last category. This option may not be specified if splits are defined on the SPLIT FILE command.

    DESCENDING. Dependent variable values are sorted in descending order, from the highest value to the lowest value.

  • If the dependent variable is a string variable, then ascending and descending order are locale-dependent.
  • If an events/trials specification is used, then the events variable must be specified first, followed by the OF keyword, and then the trials variable or number.
  • If an events/trials specification is used, then DISTRIBUTION = BINOMIAL must be specified on the MODEL subcommand. In this case, the procedure automatically computes the ratio of the events variable over the trials variable or number.
  • The events and trials variables must be numeric.
  • The events variable is usually the number of successes for each case. Data values must be nonnegative integers. Cases with invalid values are not used in the analysis.
  • If a trials variable is specified, data values must be positive integers, and each value must be greater than or equal to the corresponding events value for a case. Cases with invalid values are not used in the analysis. If a number is specified, then it must be a positive integer, and it must be greater than or equal to the events value for each case. Cases with invalid values are not used in the analysis.
  • The events and trials options are invalid if a dependent variable name is specified.
  • The names of the factors and covariates, if any, follow the dependent variable or events/trials specification. Names of factors are specified following the keyword BY. Names of covariates are specified following the keyword WITH.

    ORDER Keyword

    The ORDER specification following a list of factor variable names determines the sort order of factor values. This order is relevant for determining a factor's last level, which may be associated with a redundant parameter in the estimation algorithm.

    ASCENDING. Factor variable values are sorted in ascending order, from the lowest value to the highest value. This is the default order.

    DATA. Factor variable values are not sorted. The first value encountered in the data defines the first category; the last value encountered defines the last category. This option may not be specified if splits are defined on the SPLIT FILE command.

    DESCENDING. Factor variable values are sorted in descending order, from the highest value to the lowest value.

  • Covariates must be numeric, but factors can be numeric or string variables.
  • Each variable may be specified only once on the variable list.
  • The OFFSET and SCALEWEIGHT variables may not be specified on the GENLIN command variable list.
  • The SUBJECT and WITHINSUBJECT variables may not be specified as dependent, events, or trials variables on the GENLIN command variable list.
  • Cases with missing values on the dependent variable, the events or trials variable, or any covariate are not used in the analysis.