GLOR Subcommand (GENLOG command)
GLOR
(Generalized
Log-Odds Ratio) specifies the population contrast variable(s). For
each variable specified, GENLOG
tests the null hypothesis that the generalized log-odds ratio equals
0 and displays the Wald statistic and the confidence interval. You
can specify the level of the confidence interval using the CIN
significance-level keyword on CRITERIA
. By default, the confidence level
is 95%.
- The variable sum is 0 for the loglinear model and for each combined level of independent variables for the logit model.
- Variables specified as dependent or independent variables
in the variable list cannot be specified on
GLOR
. - The coefficient is not applied on a case-by-case basis. The weighted mean for a cell is applied to that cell.
- If multiple
GLOR
subcommands are specified, the last specification takes effect.
Example
GENLOG A B
/GLOR=COEFF
/DESIGN=A B.
- The variable COEFF contains the coefficients of two dichotomous factors A and B.
- If the weighted cell mean for COEFF is 1 when A equals B and –1 otherwise, this example tests whether the log-odds ratio equals 0, or in this case, whether variables A and B are independent.