+--------------------------------------------------------------------+
| |
| RAW_OUTPUT_FILE |
| |
+--------------------------------------------------------------------+
MEANING: NONMEM Raw Output file
CONTEXT: NONMEM Output Files
DISCUSSION:
With NONMEM 7, a NONMEM run produces several additional output files.
These provide a more efficient way of way of extracting numerical
results from the analysis than by obtaining them from the NONMEM
report file.
(See additional_output_file).
Names of the files start with "root", where root is the root name (not
including extension) of the NM-TRAN control stream file given at the
nmfe7 command line, or root="nmbayes" if the control stream file name
is not passed as an argument to the NONMEM executable. E.g., in this
example the root is myexample.
nmfe7 myexample.ctl myexample.res
The raw output file ("root.ext") will be named myexample.ext
A different name for this file may be specified using the FILE=file-
name option of the $ESTIMATION record. This filename is used for all
subsequent problems, until a new FILE=filename option is coded. The
raw output file is only generated if the $ESTIMATION record is present
and accumulates information from each Estimation step.
The first two lines of each file is the same. First is a header line
that begins with the word TABLE, such as:
TABLE NO. n:
These tables are not related to the ones produced by the $TABLE
record. The value of n is incremented each time the Estimation Step
is implemented, i.e., once per $ESTIM record, when MAXEVAL is not 0
and $EST is not omitted. The $EST records may be in the same problem,
or in subsequent problems. If $COV is present in a problem without
$EST (e.g., when $MSFI is used), then n is also incremented.
Next on the header line is the analysis text (same as given on the
#METH: line in the report file), e.g., "First Order".
In the raw output file, the analysis text is followed by the goal
function text (same as given on the #OBJT: line in the report file),
e.g.,
Goal Function=MINIMUM VALUE OF OBJECTIVE FUNCTION
At the end of the header line is:
Problem=1 Subproblem=0 Superproblem1=0 Iteration1=0
Superproblem2=0 Iteration2=0
The values of Problem, Subproblem, Superproblem1, Iteration1, Super-
problem2, Iteration2 are the same values as would be found in the fol-
lowing variables in modules during the run:
IPROB IREP S1NUM S1IT S2NUM S2IT
(See Problem_Iteration_Counters).
The second line of the raw output file contains the column headers for
the table, all on one line.
Options NOTITLE and NOLABEL and FORMAT of the $ESTIMATION record can
be used to supress the first and/or second lines, and to modify the
format of the values in the table. They apply to raw and additional
output files, but only for the current Estimation Step. The options
must be re-specified with each $EST record.
Values of the Parameters and of the objective function are printed to
the raw output file every PRINT iterations. The default order is:
ITERATION THETA, SIGMA, OMEGA, OBJ (Objective function). Values of
Sigma and Omega are in the original parameterization, as in the file
INTER. The option ORDER of the $ESTIMATION record may be used to
change the order of THETA, SIGMA, and OMEGA.
(See order_option).
The iteration number, which is the first value in every line, is typi-
cally positive, but also may be negative under the following condi-
tions:
1) The burn-in iterations of the MCMC Bayesian analysis are given
negative values, starting at -NBURN, the number of burn-in itera-
tions requested by the user. These are followed by positive
iterations of the stationary phase.
2) The stochastic iterations of the SAEM analysis are given nega-
tive values. These are followed by positive iterations of the
accumulation phase.
3) Iteration -1000000000 (negative one billion) identifies the
line that contains the final result (thetas, omegas, and sigmas,
and objective function) of the particular analysis. These are
the value in the NONMEM output file that are labelled FINAL
PARAMETER ESTIMATE. The order is theta, sigma, omega.
4) Iteration -1000000001 identifies the line that contains the
standard errors of the final population parameters. These are
the value in the NONMEM output file that are labelled STANDARD
ERROR OF ESTIMATE. The order is theta, sigma, omega. This line
is present if the Covariance Step is successful, or with
METHOD=BAYES.
5) Iteration -1000000002 identifies the line that contains the
eigenvalues of the correlation matrix of the variances of the
final parameters.
6) Iteration -1000000003 identifies the line that contains the
condition number , lowest, highest, Eigenvalues of the correla-
tion matrix of the variances of the final parameters.
7) Iteration -1000000004 identifies the line that contains the
OMEGA and SIGMA elements in standard deviation/correlation format
8) Iteration -1000000005 identifies the line that contains the
standard errors to the OMEGA and SIGMA elements in standard devi-
ation/correlation format
9) Iteration -1000000006 identifies the line that indicates 1 if
parameter was fixed in estimation, 0 otherwise.
10) Iteration -1000000007 lists termination status (first item)
followed by termination codes.
In nm73, termination status catalogs the error status:
For traditional analyses, an error number is listed. If neg-
ative, the analysis was user-interrupted
For EM/Bayes analysis, error numbers map as follows:
0,4: optimization was completed
1,5: optimization not completed (ran out of iterations)
2,6: optimization was not tested for convergence
3,7: optimization was not tested for convergence and was
user interrupted
8,12: objective function is infinite or all individual
objective fuctions are zero. problem ended
16,20: All individual objective fuctions are zero. problem
ended
4,5,6,7,12,20: reduced stochastic/stationary portion was not
completed prior to user interrupt
11) Iteration -1000000008 lists the partial derivative of the
likelihood (-1/2OFV) with respect to each estimated parameter.
This may be useful for using tests like the Lagrange multiplier
test.
See the description of root.xml in Guide INTRODUCTION TO NONMEM 7
for versions 7.4 and higher.
A file simliar to the raw output file is root.cnv. This file contains
convergence information for the Monte Carlo/EM methods, if CTYPE>0:
1) Iteration -2000000000 identifies the line that contains the
mean of last CITER values (mean objective function is that of
second to last CITER values)
2) Iteration -2000000001 identifies the line that contains the
standard deviation of last CITER values (for objective function,
STD of second to last CITER values)
3) Iteration -2000000002 identifies the line that contains the
linear regression p-value of last CITER values against iteration
number. (for objective function, p-value of second to last CITER
values)
4) Iteration -2000000003 identifies the line that contains the
Alpha used to assess statistical significance (p-value<alpha)
Please note the following: The Sigma values are in their Cholesky
format, as this is the form in which convergence of these values
are tested. The Alpha are those based on ones actually used for
convergence test of that parameter, or which would have been used
on that parameter if CTYPE were of proper type. The alpha may be
bonferoni corrected because of multiple comparisons, depending on
number of parameters that were tested or would have been tested.
Objective function alphas are not bonferoni corrected.
For importance sampling and iterative two stage, the average
objective function listed in root.cnv could be used as an alter-
native to the final objective function for likelihood ratio
tests.
REFERENCES: Guide Introduction_7
Go to main index.
Created by nmhelp2html v. 1.0 written by Niclas Jonsson (Modified by AJB 5/2006,11/2007,10/2012)