You are here

How do I subset my data prior to analysis, and how do I use legacy data where sub-setting has been specified in NONMEM code?

MDL does not support the equivalent of “IGNORE” characters or “IGNORE” and “ACCEPT” Boolean clauses to subset data prior to analysis. The data passed to the target software must be only the data to be used in estimation. It is good practice to subset data for analysis using a scripting language, rather than changing data records by hand. We suggest that users use R commands in their script to subset, filter and manipulate data prior to analysis, write the data for analysis to file and then modify the MDL Data Object to reference the appropriate source file in the SOURCE block.
For legacy data, we recommend using the “ignored” function within the “metrumrg” R package (https://r-forge.r-project.org/projects/metrumrg/). This function reads a NONMEM control stream and creates a TRUE/FALSE logical flag for whether the records meet IGNORE and/or ACCEPT criteria specified in the NMTRAN code. This will allow the user to identify which data records have been dropped by NONMEM. Filtering on the original data based on these criteria will allow them to create a dataset ready for analysis with MDL.