"Tip of the Week #7 - April 8, 2002" - subsetting your data 'on the fly'

-----Original Message----- From: Prins, N.H. (Klaas) Sent: Friday, March 08, 2002 10:21 AM To: 'nmusers@globomaxnm.com' Subject: [NMusers] EVID data item and beyond Dear nm-users, Currently I am working on a study that consisted of 4 different study parts in which several dosing schemes were tested, i.e. 3 schemes of multiple dosing and 1 of continuous infusion. Initially, I created separate data sets for each study part and inserted intermediate time points to get smooth curves when plotting observed and predicted versus time. Using either the MDV (0=observation, 1=missing) or EVID (0=observation, 1=dosing, 2=other/simulation) data item, I managed to get decent fits for each study part. Also combining data from different study parts resulted in succesful NM runs. However, I did not feel satisfied with making more than one dataset and a couple of control streams. Therefore, I set out to put together a single dataset containing data of all study parts. Then, the aim was to flag out the study parts that did not need to be evaluated in specific run, by setting EVID to 2 (other event). In this manner, I created 6 different evaluation schemes and listed them in data items EV1 to EV6. For instance, scheme EV4 was created to evaluate study parts 1 and 4 only, then all records for those parts were set to 0, 1 and 2 (see above) and all other records were set to 2 (flagged as 'other'). The $INPUT control stream read the following: $INPUT ID TIME AMT RATE DV SEX WT PART EV1=DROP EV2=DROP EV3=DROP EV4=EVID EV5=DROP EV6=DROP So, by toggling on and off the EVx=EVID/DROP, I hoped to get my desired evaluation. It did not work since the nm-tran reported for the data records of study parts that were flagged out: THE DV DATA ITEM IS POSITIVE, BUT THE MDV DATA ITEM IS 1 I did not specify a MDV data item, so this must refer to a 'generated MDV'. I do understand this, since for these parts, the observations are read and NM is suprised that it is asked to consider this as a simulation record. Now, after this exhausting problem description, could somebody comment on this issue and provide hints on how to surpass this? Any suggestion is much appreciated. Thanks in advance and have a nice weekend, Klaas Prins -----END of Original Message----- Actually, Klaas was very close to the solution of his problem. The issue Klaas encountered was that not only do you need to "pre-specify" the EVID for NONMEM, you also need to pre-specify the MDV and ANY DOSING-RELATED data items such as AMT and RATE because this scheme relies on: 1. pre-defining to NM-TRAN those records which you choose not to analyze as MDV=1 (missing) and that DV is not expected to be positive for MDV=1 2. other data item records (EVID=2) may not specify dosing information. To circumenvent these limitations one must apriori create columns in your data representing all the combinations of data you wish to subset for analysis by specifying the EVID, MDV, AMT (and RATE if needed). I have created an example data file, test.csv, in which I make four subsets of the full data set: subset 0: all of the data, specified by data columns EV0, MV0, AM0 subset 1: subjects 1-4, specified by data columns EV1, MV1, AM1 subset 2: subjects 5-8, specified by data columns EV2, MV2, AM2 subset 3: subjects 9-12, specified by data columns EV3, MV3, AM3 subset 4: subjects 1-4 & 9-12, specified by data columns EV4, MV4, AM4 (This will, of course, increase the number of columns in your original datafile, but NOT in the number of columns in the datafile used by NONMEM, FDATA, becuause the columns not used in a given run will be excluded using "DROP=" and because NM-TRAN will ordinarily add columns for EVID, MDV and AMT anyway.) The datafile for one subject looks like: C,Data Desc: Theophylline PK Study,,,,,,,,,,,,,,,,,, C,ID,TIME,DV,WT,AM0,AM1,AM2,AM3,AM4,EV0,EV1,EV2,EV3,EV4,MV0,MV1,MV2,MV3,MV4 ,1,0,.,79.6,4.02,4.02,0,0,4.02,1,1,2,2,1,1,1,1,1,1 ,1,0,0.74,.,.,.,.,.,.,0,0,2,2,0,0,0,1,1,0 ,1,0.25,2.84,.,.,.,.,.,.,0,0,2,2,0,0,0,1,1,0 ,1,0.57,6.57,.,.,.,.,.,.,0,0,2,2,0,0,0,1,1,0 ,1,1.12,10.5,.,.,.,.,.,.,0,0,2,2,0,0,0,1,1,0 ,1,2.02,9.66,.,.,.,.,.,.,0,0,2,2,0,0,0,1,1,0 ,1,3.82,8.58,.,.,.,.,.,.,0,0,2,2,0,0,0,1,1,0 ,1,5.1,8.36,.,.,.,.,.,.,0,0,2,2,0,0,0,1,1,0 ,1,7.03,7.47,.,.,.,.,.,.,0,0,2,2,0,0,0,1,1,0 ,1,9.05,6.89,.,.,.,.,.,.,0,0,2,2,0,0,0,1,1,0 ,1,12.12,5.94,.,.,.,.,.,.,0,0,2,2,0,0,0,1,1,0 ,1,24.37,3.28,.,.,.,.,.,.,0,0,2,2,0,0,0,1,1,0 To analyze all to the data, specify EVID=EV0, MDV=MV0, AMT=AM0 in $DATA of your control stream (and DROP= is specified for the remaining columns), e.g.: ;Model Desc: NONMEM & PDx-Pop compliant control stream ;Project Name: TIPS ;Project ID: NO PROJECT DESCRIPTION $PROB RUN# 011 DATA $INPUT C ID TIME DV WT AMT=AM0 DROP=AM1 DROP=AM2 DROP=AM3 DROP=AM4 EVID=EV0 DROP=EV1 DROP=EV2 DROP=EV3 DROP=EV4 MDV=MV0 DROP=MV1 DROP=MV2 DROP=MV3 DROP=MV4 $DATA test.CSV IGNORE=C $SUBROUTINES ADVAN2 $PK KA=THETA(1)*EXP(ETA(1)) K =THETA(2)*EXP(ETA(2)) CL=THETA(3)*EXP(ETA(3)) SC=CL/K $THETA (0,3) (0,.08) (0,.04) $OMEGA 0.1 ;[p] 0.1 ;[p] 0.1 ;[p] $ERROR Y=F+EPS(1) $SIGMA .1 ;[A] $EST MAXEVAL=450 PRINT=5 MSF=011.MSF $COV $TABLE ID AMT TIME WT EVID MDV FILE=011.TAB ONEHEADER NOPRINT To analyze subset 1 of the data, specify EVID=EV1, MDV=MV1, AMT=AM1 in $DATA of your control stream (and DROP= is specified for the remaining columns), e.g.: $INPUT C ID TIME DV WT DROP=AM0 AMT=AM1 DROP=AM2 DROP=AM3 DROP=AM4 DROP=EV0 EVID=EV1 DROP=EV2 DROP=EV3 DROP=EV4 DROP=MV0 MDV=MV1 DROP=MV2 DROP=MV3 DROP=MV4 The control stream, tip7.ctl, and the complete datafile, test.csv, for this example may be found in the NONMEM Repository @ GloboMax: ftp://ftp.globomaxnm.com/Public/nonmem/tips/ [Generic Disclaimer: Verify that this code or any other code you receive from an outside source works with YOUR DATA.] ***************************************************************** If you have a "tip" or a better way to do things, by all means, feel free to post them! One of the reasons for doing this (other than good PR for GloboMax), is to stimulate discussion. We at GloboMax can learn from your experience as well. Do you have a "tip" you would like to share, but would prefer to remain anonymous? If so, you may forward it to us and your identity will be withheld. Distribution as a "Tip of the Week" will be at the discretion of nmconsult@globomaxnm.com. ***************************************************************** nmconsult@globomaxnm.com GloboMax LLC 7250 Parkway Drive, Suite 430 Hanover, MD 21076 Voice: (410) 782-2205 FAX: (410) 712-0737