One of the features of Emacs Speaks NONMEM is the automatic generation of certain pieces of code that may be useful for analysis, such as conditional weighted residuals, splitting up control streams to bypass the 20 (<NONMEM 6.2) or 50 variable PRED generated output variable limitation of NONMEM, and automatically generating code secondary parameters. Some of the code is specific to the NONMEM version that is used. For this reason a discussion of how Emacs Speaks NONMEM determines the version is included.

Conditional Weighted Residuals

Conditional Weighted Residuals were proposed for NONMEM by Andrew Hooker. Instructions on what to do to generate the appropriate piece of code are found in Xpose's website. Emacs Speaks NONMEM helps with the generation by allowing

  • Appropriate Code to be generated depending on the version of NONMEM selected
  • Code to be reflective of the estimation method using. Currently, the default method where conditional weighted residuals are generated is the first order conditional estimation method. Other methods may be specified through the menu:
    • Automation
      • Options

If you plan to use Perl Speaks NONMEM, you will need to select

  • Automation
    • Options
      • PsN compatibility — Use ../../ for .est files

This allows the .est files required for the calculation of conditional weighted residuals to reside in the same directory as the rest of the files and the control stream.

Large Tables

When there are over 20 PRED generated variables (or 50 in NONMEM VI.2 or greater), NONMEM can not output the variables using one control stream. One solution is to output the model specification, and use the same model to output the missing variables. Last the variables should be merged together. Emacs Speaks NONMEM can do this automatically when you save, and open as if there were no limit to the number of variables that can be output. When there are more variables output than can be interpreted by NONMEM, Emacs Speaks NONMEM modifies and splits the files as follows:

  • The open control stream is stripped of variables over the 20/50 PRED generated variable limit, and places comments at the end of the control stream that allows NONMEM to know about these variables and open the base control stream as if this translation did not occur.
  • Change the table names in this file to have a trailing ~0 in the file name. For example a table output of patab1 would be changed to patab1~0
  • Create the appropriate number of control streams that read in the model specification file and output variables that are missing from patab1~0. For example,
    • For the first new control stream, run1.mod, create a new control stream run1~1.mod that outputs tables where the variables are missing from the table specified in run1.mod. For a table patab1, this control stream would output patab1~1.
    • More control stream and table information are created if necessary. (i.e. run1~2.mod, patab1~2, run1~3.mod, patab1~3, etc.
    • To produce the final table, you will need to:
      • Run the first control stream run1.mod. Once this has been completed, run all subsequent control streams run1~2.mod, run1~3.mod, etc.
      • Merge the output datasets together to the final tables.

Note that you will have to create a script to paste the tables together line-by-line. An example of this using Xpose and R, is as follows:

xpose.merge <- function(runno,tab.suffix="",
                        table.names=c("sdtab", "mutab", "patab",
                        "catab", "cotab", "mytab", "extra", "xptab", "cwtab"),
              "cwtab")) {
    tab.files <- sapply(paste(table.names, runno, tab.suffix,"~*",sep = ""),Sys.glob);
    for (i in 1:length(tab.files)){
        merged.file <- NULL;
        out.file.times <- c();
        if (length(tab.files[[i]]) >= 1){
            for (file in tab.files[[i]]){
                out.file.times <- c(out.file.times,$mtime);
                con <- file(file,"r");
                cont <- readLines(con);
                if (is.null(merged.file)){
                    merged.file <- cont;
                } else {
                    cont[1] <- "";
                    merged.file <- paste(merged.file,cont);
            write.file <- TRUE;
            file <- gsub("~[0-9]+$","",tab.files[[i]][1]);
            if (file.exists(file)){
                file.time <-$mtime;
                if (all(file.time > out.file.times)){
                    cat("Not writing ",file," since it is newer than the  files ",file,"~*.\n",sep="");
                    write.file <- FALSE;
            if (write.file){
                merged.file <- paste(merged.file,collapse="\n");

After loading this function, the tables can be merged with xpose.merge(1).

This automatic splitting and reassembling can be toggled by:

    • Automatically...
      • Generate Code
        • Create multiple control streams to output more generated variables than NONMEM can handle

Secondary Parameters Output

Another option that Emacs Speaks NONMEM allows is the generation of secondary parameters for specific PK ADVAN/TRANS combinations. This creates a section of code that generates useful secondary parameters such as half-life when Emacs Speaks NONMEM can determine that the parameter calculation applies.

For example, in a one-compartment model, Emacs Speaks NONMEM adds the following parameters:

  CL   = THETA(1)*DEXP(ETA(1))
  V    = THETA(2)*DEXP(ETA(2))

  ;| Begin Secondary Parameters
    ;; Fixed Amount Variable
    DOSE = AMT
    ;; Time After Dose Calculation
    TAD  = 0.0
    TAD  = TIME - TDOS

  K    = CL/V
  HLK  = -DLOG(0.5)/K
  ;| End Secondary Parameters

The code between ;| Begin Secondary Parameters and ;| End Secondary Parameters will be auto-generated based on the ADVAN/TRANS combination that is used. This option may be toggled at:

    • Automatically...
      • Generate Code
        • On save, generate derived paramters like duration, AUC, etc., when known.

Determining NONMEM version

Emacs Speaks NONMEM determines the NONMEM version in the following way:

  • The version of NONMEM is specified within the header (must be prompted for, and setup appropriately; see Configure header).
  • If Emacs Speaks NONMEM cannot determine the version it assumes the version specified in the Use menu.