BUG IN ETA-SHRINKAGE CALCULATION
1. For and eta describing IIV/BSV, the calculation of
shinkage is correct. However, if the eta represents
IOV/BOV, I think the same approach is used as for IIV,
which means that only the first-occasion eta for each
individual counts. If there is some systematic
difference between occasion, this calculation would
indicate more shrinkage than correct and it certainly
would be more precise to instead do the following:
Create a table with all unique etas within an
individual and use these etas instead (would change
nothing in the IIV-etas but IOV-etas would be used
correctly). The creation of this table probably has to
be done outside of NONMEM. See bug in this example:
/users/jakob/LASSO/prazosin_data/bug_shrinkage
Log in to post a comment.
