What causes an error that either the COVB= data set is missing a variable or that the matrix is not symmetric?


In PROC MIANALYZE, the COVB= data set is only needed when you specify the MULT option in either the PROC MIANALYZE statement or in the TEST statement. Otherwise, it is recommended that you remove the COVB= option.

If you specify the COVB= and MULT options and receive an error message that either a variable is not found or that your covariance matrix is not symmetric, the most common cause for this is that the EFFECTVAR= suboption of the COVB= option has not been properly assigned. See the table below for a listing of each of the EFFECTVAR= values and the corresponding procedures.

EFFECTVAR=Valid Procedures
STACKINGGENMOD
LIFEREG
LOGISTIC
PHREG
PROBIT
QUANTREG
ROBUSTREG
SURVEYLOGISTIC
ROWCOLGLIMMIX
MIXED