Hi All, I’m finishing up an NLME analysis and using bootstrapping for validation. I hve two questions about the bootstrapping technique used in Phoenix NLME. 1. How are the replicates selected. I have about 28 subjects with some 200 or so concentrations. With resampling, the original dataset is “resampled” by leaving out some of the data from the original data to form a replica set. FDA and literature sources suggest using 200 replicas for bootstrapping in PK NLME work. Okay, does the Phoenix NLME application of bootstrapping drop samples or whole subjects from the original data to form the replica data and how are they chosen? 2. As bootstrapping does take a bit of time, how many cores can Phoenix NLME use productively. I’m running on only one core right now as the multicore option was not installed with Phoenix NLME was installed. I will get it reinstalled with the multicore option, but I am also hoping to get a 4 core MPU with the new laptop. Can Phoenix NLME use all 4 cores or can it even use more. Run time for the bootstrapping program with 28 subjects and a 2-cmpt model is taking about 10 hours. I would really like to cut that down quite a bit. Can 2 cores cut it back to 5 hours, and 4 cores back to 2.5 hours? Any advice would be appreciated. Jim McDowell
Hi Jim, we’re looking into your other problem with SE calcs but I wanted to get back to you on these two questions as I understand them 1) Regarding on how the replicates are chosen; the data is replicated at the subject level, not the individual observation level. So an original data set with Subjects 1, 2,3,4,5,6,7,8,9,10 might be replicated to 1a,1b,2,3,4,5,7a,7b,9,10 where 1a and 1b are exact copies of the original subject 1’s data except for the subject id, and similarly 7a and 7b are exact copies of the original subject 7’s data. In the replicated data set, some subjects are left out, and some appear more than once, but the total number of subjects is kept the same. 2) Phoenix parallelizes single runs, and indeed 2 processors should run significantly faster than 1 processor, although not necessary twice as fast. Phoenix can utilise as many cores as are available, or you can choose to specify use 3 out of the 4 you have; 
Further more if you have access to a cluster 6.2 has an improved Job Managment Scheduler to submit jobs to run on remote machines.