Following PK data analysis I will have the NCA values for PK parameters from men and women subjects e.g. Cmax values from treatments with 3 different formulations (A,B and C). There will be 3 periods, 3 treatments and it will be a 3 way cross-over design. So I organize and the C max values according to gender and treatment. Can one test for the statistical difference in mean values of a parameter like Cmax according to formulation treatment group and gender Phoenix 6.3. I am exploring the statistical capabilities of the program. Angus
Hi Angus, I’m sure you can. 
Have a column denoting gender (0/1, f/m, whateveryoulike…) and map it in the BE setup as “Classification”. Model Specification (Fixed Effects): sequence+period+treatment+gender+treatment*gender Example from one of my studies (6×3 Williams’ design): Partial Tests of Model Effects Hypothesis Numer_DF Denom_DF F_stat P_value ------------------------------------------------------------------------ int 1 5 268.473 0.0000 sequence 5 5 1.03084 0.4871 period 2 18 2.75785 0.0902 treatment 2 18 92.1666 0.0000 gender 1 5 0.645225 0.4583 gender*treatment 2 18 1.4161 0.2684 This was a dose-proportionality study – so don’t worry about about the highly significant treatment differences. PS: @ the moderators: Can you move this thread to the PHX board?
Helmut: Thanks! I am exploring Phoenix BE for the first time. I see the treatment effect and note the Williams balanced design {6 sequences}. When I was in the VC industry we did a 3 way cross-over study on carbamazepine, but it was not balanced {only 3 sequences}…“Schadenfreude”. It was a very experimental formulation study not at all pivotal to the submission. I wonder if it was less expensive to run that way compared with Williams design. I will see if I still have that data. Angus