https://communities.sas.com/t5/Statistical-Procedures/2-way-random-effects-ANOVA/m-p/76486#M3655 Hi everyone.<BR /> <BR /> May I please ask what this " Convergence criteria met but final<BR /> hessian is not positive definite" outputed by proc mixed means?<BR /> <BR /> my code is<BR /> proc mixed data=temp;<BR /> class Subject factory;<BR /> model log_exp = / solution;<BR /> random int / subject=Subject solution;<BR /> random int / subject=factory solution;<BR /> random int / subject=factory *Subject solution;<BR /> run;<BR /> <BR /> Many thanks in advance. Thu, 22 Oct 2009 10:11:11 GMT Miracle 2009-10-22T10:11:11Z https://communities.sas.com/t5/Statistical-Procedures/2-way-random-effects-ANOVA/m-p/76486#M3655 Hi everyone.<BR /> <BR /> May I please ask what this " Convergence criteria met but final<BR /> hessian is not positive definite" outputed by proc mixed means?<BR /> <BR /> my code is<BR /> proc mixed data=temp;<BR /> class Subject factory;<BR /> model log_exp = / solution;<BR /> random int / subject=Subject solution;<BR /> random int / subject=factory solution;<BR /> random int / subject=factory *Subject solution;<BR /> run;<BR /> <BR /> Many thanks in advance. Thu, 22 Oct 2009 10:11:11 GMT https://communities.sas.com/t5/Statistical-Procedures/2-way-random-effects-ANOVA/m-p/76486#M3655 Miracle 2009-10-22T10:11:11Z https://communities.sas.com/t5/Statistical-Procedures/2-way-random-effects-ANOVA/m-p/76487#M3656 Usually, this means the calculations came up with a negative value for one of the variance components. Take a look at the Covariance Parameter Estimates, and you should see one set to 0. You may need to adjust your model accordingly to avoid this situation. Thu, 22 Oct 2009 18:16:53 GMT https://communities.sas.com/t5/Statistical-Procedures/2-way-random-effects-ANOVA/m-p/76487#M3656 sfleming 2009-10-22T18:16:53Z