Re: FDATA file bug
Hi Alison,
Thank you so much for your comprehensive suggestion, which is definitely
helpful. This topic is like that we are still driving a manual transmission
car in an era of self-driving car. Why could not NONMEM be updated to make
the life easier for the end-users? A simple csv file with headings can save
us a lot of time.
I still think that check of FDATA is the most intuitive and direct way to
ensure correct input. Sometimes, plots may not sufficiently provide details
for every data point, especially when you have thousandths of points
cluttered in one plot. Additionally, graphic evaluation can be highly
dependent on the visual sensitivity of the investigator.
Joel S. Owen and Jill Fiedler-Kelly also suggested examination of the FDATA
file in the book "Introduction to Population Pharmacokinetic/
Pharmacodynamic analysis with nonlinear mixed effects models". On page 33,
it says "Such exclusions from the datafile can be confirmed by viewing the
FDATA output file"
I followed your suggestion but got errors for the proposed model file as
listed below.
$PROBLEM INTRAVENOUS BOLUS STUDY
$DATA CS1_IV1EST_PAR.CSV IGNORE=C CHECKOUT
$INPUT ID TIME DV AMT DOSE MDV AGE WT SCR ISM CLCR AA BB CC DD EE FF
$SUBROUTINE ADVAN1 TRANS2
$PK
CL = THETA(1)*EXP(ETA(1))
V= THETA(2)*EXP(ETA(2))
S1 = V
$ERROR
IPRED = F
Y=F+F*ERR(1)+ERR(2)
$THETA
(0.1,1)
(1,10,20)
$OMEGA 0.09
0.09
$SIGMA 0.09
1
$TABLE ID TIME DV AMT DOSE MDV AGE WT SCR ISM CLCR AA BB CC DD EE FF
NOAPPEND NOPRINT FILE=test.tab
LFORMAT="(' ',20A6)"
RFORMAT="(20F6.1)"
The ERROR message is "There was an error when running nmfe, NMtran could
not be initiated for model 1"
Thank you,
Mark
Quoted reply history
On Thu, Dec 5, 2019 at 12:14 PM Alison Boeckmann <
[email protected]> wrote:
> Hi Richard and others:
> Thanks for explaining FDATA. There is no bug.
> FDATA would have a simpler, space delimited format if there were fewer
> characters per line.
> It only uses a dense format when it has not got enough room to keep the
> length of lines in FDATA to 80.
> If you add "WIDE" to the $DATA record, then FDATA will have spaces between
> data items and will be easier to read.
>
> Mark wrote: "I want to confirm the right input dataset into NONMEM. "
>
> I see no reason to look into NONMEM internal files such as FDATA.
> The best way to check the data was suggested by Lewis Sheiner in NONMEM
> Users Guide V, Introductory Guide,
> Chapter 11 Model Building, section 3 *3 *3. Check-out — Index Plots.
>
> I took the control stream and data file from Mark's email and made three
> changes
> to the control stream, following the spirit of Sheiner's suggestions.
>
> Add CHECKOUT to the $DATA record. This tells NONMEM not to try to compute
> the objective function.
> It may not be able to do this if the data or model is faulty.
>
> Delete the $ESTIM record. Why waste computer time and risk another failure
> if the data is faulty?
>
> Sheiner suggested index plots, but right now we are just checking the
> basic set up rather than
> hunting for errors in the data.
>
> Replace $TABLE with the following. I would not ordinarily use
> LFORMAT/RFORMAT
> but I put them in so as to generate something scrunched that I can copy
> and paste into this email:
>
> $TABLE ID TIME DV AMT DOSE MDV AGE WT SCR ISM CLCR AA BB CC DD EE FF
> NOAPPEND NOPRINT FILE=test.tab
> LFORMAT="(' ',20A6)"
> RFORMAT="(20F6.1)"
>
> When I run this, test.tab is generated.
>
> ID TIME DV AMT DOSE MDV AGE WT SCR ISM CLCR AA
> BB CC DD EE FF
> 1.0 0.0 0.0 100.0 100.0 1.0 34.8 38.2 1.1 0.0 42.6 42.6
> 42.6 42.6 42.6 42.6 42.6
>
> Its easy to see that this matches the original data.
>
> 1, 0 , 0, 100, 100, 1, 34.823, 38.212,
> 1.1129, 0 , 42.635, 42.635, 42.635, 42.635, 42.635, 42.635, 42.635
>
> With WIDE option of $DATA, the table is unchanged, though FDATA is easier
> to examine visually and
> the format in FCON is simper.
>
> Hope this helps,
> Alison
>
> On Wed, Dec 4, 2019, at 11:54 PM, Rikard Nordgren wrote:
>
> Hi Mark,
>
> I tested your model and dataset and checked the FORM section of the FCON
> file:
>
> FORM
>
> (E3.0,E4.0,E8.0,2E3.0,E1.0,2E6.0,E7.0,E1.0,5E6.0/2E6.0,1F2.0)
> and the FDATA, first record here:
> 1 0 0100100134.82338.212 1.1129042.63542.63542.63542.63542.635
> 42.63542.635 1
>
> It seems to be fine. Please refer to how fortran uses the format statement
> to parse strings using field sizes rather than delimiters.
>
> Best regards,
> Rikard
>
>
>
>
>
> On 2019-12-05 08:23, Mark Tepeck wrote:
>
> Hi Nordgren and Denny,
>
> Thanks for your tips. Here are the NONMEM script, input CSV file row 1,
> and FDATA row 1. I also attached the original model file, CSV input file as
> well as the FDATA (not sure NMuser allows attachment) for troubleshooting.
> I did observe that multiple columns from input were combined to fewer
> columns in the FDATA file. I also found occasional generation of *fdata-1*
> file but not sure its role.
>
> *The input CSV file (17 columns) *
> CID, TIME, CONC, AMT, DOSE, MDV, AGE, WT, SCR, ISM, CLCR,
> AA, BB, CC , DD, EE, FF
> 1, 0 , 0, 100, 100, 1, 34.823, 38.212,
> 1.1129, 0 , 42.635, 42.635, 42.635, 42.635, 42.635, 42.635, 42.635
>
> *FDATA (6 columns):* 1, 0, 0100100134.82338.212,
> 1.1129042.63542.63542.63542.63542.635, 42.63542.635,1
>
> *NONMEM script:*
>
> $PROBLEM Test Fdata
> $DATA CS1_IV1EST_PAR.CSV IGNORE=C
> $INPUT ID TIME DV AMT DOSE MDV AGE WT SCR ISM CLCR AA BB CC DD EE FF
> $SUBROUTINE ADVAN1 TRANS2
> $PK
> CL = THETA(1)*EXP(ETA(1))
> V= THETA(2)*EXP(ETA(2))
> S1 = V
>
> $ERROR
> IPRED = F
> Y=F+F*ERR(1)+ERR(2)
>
> $THETA
> (0.1,1)
> (1,10,20)
>
> $OMEGA
> 0.09
> 0.09
>
> $SIGMA
> 0.09
> 1
> $ESTIMATION METH=1 MAXEVAL=9999 PRINT=5 POSTHOC
> $TABLE ID TIME DV FILE=CS1_IV1ESTFPDF.fit
>
>
> I really appreciate your input.
>
> Mark
>
>
>
>
>
> On Tue, Dec 3, 2019 at 11:57 PM Rikard Nordgren <
> [email protected]> wrote:
>
> Hi,
>
> NM-TRAN will not add columns not listed in $INPUT or dropped/skipped to
> FDATA. Could that be the reason?
>
> Best regards
> Rikard Nordgren
>
>
>
> On 2019-12-04 01:27, Mark Tepeck wrote:
>
> Hi All,
>
> I am using NONMEM 7.4.3. I want to confirm the right input dataset into
> NONMEM. However, I found that the FDATA file can save only around the
> beginning 9 columns of the original CSV file. The ending columns (e.g.,
> column 10-20) are either missing or combined into a single column in the
> FDATA file.
>
> I am not sure other people encountered the same problem.
>
> Thank you,
>
> Mark
>
>
> --
> Rikard Nordgren
> Systems developer
>
> Dept of Pharmaceutical Biosciences
> Faculty of Pharmacy
> Uppsala University
> Box 591
> 75124 Uppsala
>
> Phone: +46 18
> 4714308www.farmbio.uu.se/research/researchgroups/pharmacometrics/
>
>
>
>
>
>
>
>
>
> När du har kontakt med oss på Uppsala universitet med e-post så innebär
> det att vi behandlar dina personuppgifter. För att läsa mer om hur vi gör
> det kan du läsa här: http://www.uu.se/om-uu/dataskydd-personuppgifter/
>
> E-mailing Uppsala University means that we will process your personal
> data. For more information on how this is performed, please read here:
> http://www.uu.se/en/about-uu/data-protection-policy
>
>
> --
> Rikard Nordgren
> Systems developer
>
> Dept of Pharmaceutical Biosciences
> Faculty of Pharmacy
> Uppsala University
> Box 591
> 75124 Uppsala
>
> Phone: +46 18
> 4714308www.farmbio.uu.se/research/researchgroups/pharmacometrics/
>
>
> --
> Alison Boeckmann
> [email protected]
>
>
>