NONMEM Level 2.1 Update

Dr. Grevel asks whether the updates distributed in Dec. 1994 should be expected to increase run times. The answer is no, at least not in any uniform way. Maybe there are some problems which will run faster with the updates. I don't think so, but I haven't thought very hard about this detail before writing this quick note. Dr. Grevel also seemed to indicate that he didn't receive a diskette with the updates along with Newsletter #11. He should have. We tried to send each contact person who has diskettes listed as his/her distribution medium a diskette. If you haven't received a diskette, but think you have diskettes listed as your distrubution medium, please contact the NONMEM consultant.

I finally got around to updating my NONMEM Version IV Level 2.0 to 2.1, PREDPP Version III Level 1.0 to 1.1, and NMTRAN Version II Level 2.0 to 2.1. I would have liked to use updated files from a disk but for some reason I did not get one with the Newsletter #11 of November 1994. I used the Fortran Workbench editor of Microsoft Fortran Powerstation. It went pretty smoothly. I compiled all altered files with the following statement: ...nmiv\run> fl32 /c /G4 /Ox /Op *.for There were the ususal warnings which I ignored but no compilation errors. I tested the udates with a problem which I ran back in December: ADVAN2 TRANS2, 1284 data records, and 158 subjects. I was surprised to see the iterations take a faster path to exactly the same minimum as before. The final parameter estimates were the same for all THETA's, ETA's, and EPSYLON's. The standard errors were, however, slightly different (mostly in the third digit, sometimes in the second). Also the correlations were slightly different. Is the faster (steeper) descent a common phenomenon with these updates? I would welcome that.