GENMOD parametrization and TYPE3/"Joing Table" interpretation

[JLNH]

I am running  GENMOD with REPEATED  statement and depending on whether I use /PARAM=EFFECT or PARAM=REF on the CLASS statement I am getting contradictory conclusions form the p value of Chi Sq on the TYPE3/"Join Table" . Which PARM should I select to test a particular effect?  

[sld]

The documentation may provide the answer to your question: The GENMOD Procedure: Type 3 Analysis. Among other things, it depends upon whether you have missing cells, whether you have additional factors in the MODEL statement, whether you have interactions, etc.

[StatDave]

See the bottom half of the "Details: Type 3 Analysis" section of the GENMOD documentation which explains the distinction between the Type 3 tests provided with GLM parameterization and joint tests provided with full-rank parameterizations.

[JLNH]

Thank you, I had already read it  before I posted  y question