Bioequivalence and Bioavailability Forum

Dear Groupmates,

I have faced one problem when i was analysing BE data with proc glm and proc mixed.

for one PK parameter, proc mixed showing the following msg in LOG:

"matrix G is Non positive definite and hence g-inverse has been used"
beacause of this msg the df is also showing 8 while it is a pilot study with 6 subjects. The 90% CI obtained through this two different procedure are very different.

the winnonlin results are matching with proc glm.

Please suggest some readings and more clarification on this.

which results should be reported????

Dipesh