RE: [NMusers] Data checking macro

-------- Original Message --------
Subject: RE: [NMusers] Data checking macro
From: "Bill Bachman" <bachmanw@comcast.net>
Date: Thu, September 13, 2007 4:35 pm
To: "'Mark Sale - Next Level Solutions'" <mark@nextlevelsolns.com>
Cc: <nmusers@globomaxnm.com>

Mark,

 

Ok with me to add any of those suggested plots!  I may add your covariate histograms to our scripts!  I will probably add statistical summaries of all data items to the R and S-Plus scripts (a simple one liner).

 

Perhaps some type of plots by time afte! r dose (TAD) plots may have helped in your situation?  We have automated a variety of TAD plots for the output table files (but not for the data checkout plots yet).

 

Bill

 

---

Fr! om: owner-nmusers@globomaxnm.com [mailto:owner-nmusers@globomaxnm.com] On Behalf Of Mark Sale - Next Level Solutions
Sent: Thursday, September 13, 2007 4:13 PM
Cc: nmusers@globomaxnm.com
Subject: RE: [NMusers] Data checking macro

 

Thanks Bill,
 I'll consider adding those (if Icon doesn't object). But, my experience is that errors in DV (value) are not too hard to find.  My single biggest time-wasting exercise (about 3 weeks) involved a very complex dosing scheme, several drugs, from several studies, several files, >! ; 1000 subjects (so individual plots by ID was not very practical) with an II of 12 when it should have been 24, due to a slight difference in how the source data file was formatted.  The interdose interval (and the last dose-to-dv interval) would have pickup it up.

Mark


Mark Sale MD
Next Level Solutions, LLC
www.NextLevelSolns.com
919-846-9185

-------- Original Message --------
Subject: RE: [NMusers] Data checking macro
From: "Bill Bachman" <bachmanw@comcast.net>
Date: Thu, September 13, 2007 3:48 pm
To: "'Mark Sale - Next Level Solutions'" ! <mark@nextlevelsolns.com>
Cc: <nmusers@globomaxnm.com>

PDx-Pop 2.2 (and below) has an Excel macro that imports the data sets and automatically gives the following plots:

DV vs ID, AMT (dose) vs ID, TIME vs ID, and DV vs TIME.  The idea is to graphically look for outliers in these plots.

 

PDx-Pop 3.0 (not yet released) also gives you the option to import the data into R or S-Plu! s (your choice) and automatically create the above plots and additionally individual p! lots of the DV vs TIME by ID.

 

Bill

 

---

From: owner-nmusers@globomaxnm.com [mailto:o! wner-nmusers@globomaxnm.com] On Be half Of Mark Sale - Next Level Solutions
Sent: Thursday, September 13, 2007 3:27 PM
Cc: nmusers@globomaxnm.com
Subject: [NMusers] Data checking macro

 

Colleagues,
  I suspect I'm not the only one who has, over the years had the experience of spending a week (or more) on an analysis only to find important e! rrors in the data set.  I'm hoping for some feedback on what people do to try to find these errors (short of spending a week on an incorrect data set).  To start the discussion, I've put on the Next Level web site (www.NextLevelSol! ns.com/downloads) an Excel macro that I've used, with some success to find errors.  My experience is that most errors, at least those that are hard to find, are in the dosing specification.  This macro makes histograms of:Each covariate
DVs
Dose Amts
Dose Times (after expanding the ADDL doses)
Interdose interval (after expanding the ADDL doses) - time from each to the previous
Dose to DV time (after expanding the ADDL doses) - time from each observation to the previous dose

Currently this macro is limited to 12 covariates, could be increased easily if there is interest.  It also isn't CMT specific, that is all doses are just listed, without regard to CMT, same with DVs.  I might! fix this someday.


As usual, this is an entirely in my o wn self-interest, looking for better ways to find problems in data sets, so please give feedback or ideas.


Mark