[DDI-users] DDI-users Digest, Vol 105, Issue 14

[Hoyle, Larry]

Bob-
I think that while this was a problem in DDI3.1,  DDI3.2 does finally have the objects necessary to document missing values from SAS for continuous variables. We just need the tools to catch up to using the added features (In the VariableRepresentation MissingValuesReference to a ManagedMissingValuesRepresentation)
The ManagedMissingValuesRepresentation can point to a Codelist for your
.r = 'respondent refused'
.d = "don't know"
.i  = "imputed deletion"
.  = "missing"  /* just missing from the qx or biological sample */
.l = "lost to followup"
.m = "missing portion of derived var"




--- Larry Hoyle


From: ddi-users-bounces at icpsr.umich.edu [mailto:ddi-users-bounces at icpsr.umich.edu] On Behalf Of Bob McConnaughey
Sent: Thursday, June 26, 2014 2:12 PM
To: ddi-users at icpsr.umich.edu
Subject: Re: [DDI-users] DDI-users Digest, Vol 105, Issue 14

Just fyi:
and i hope i'm not trying everybody's patience here.
Our datasets are generally a mix of demographic/behavioral variables + host of "biologic" variables whether hormonal profiles of urine samples (daily urines for up to 6 months in the study I've worked on for decades) or now, for better or worse, oodles (the scientific term..) of genetic based vars - there could be, literally, 1000s of snps for each subject. yuck. AND then  (far too many imo ;-) variables that attempt to capture all the various and sundry "exposure" variables.
    we've working towards  a more or less standard set of SAS missing values that attempt to deal with basic "missing situations" - but with the long history of some of these projects, even now we're not completely there, yet.
.r = 'respondent refused'
.d = "don't know"
.i  = "imputed deletion"
.  = "missing"  /* just missing from the qx or biological sample */
.l = "lost to followup"
.m = "missing portion of derived var"
 /* say a derived  rate of change var requires 5 days of progesterone     */
  /* hormone and only has 3 days in the sequence, .m could reflect that  */
(the above is not completely accurate - i'm at home w/ the world cup.
Character variables are different kettle of fish.

a simple study might be one i've worked on forever: Early Pregnancy Study
http://www.niehs.nih.gov/research/atniehs/labs/epi/studies/eps/index.cfm
a somewhat more complex study: Norway Facial Clefts
http://www.niehs.nih.gov/research/atniehs/labs/epi/studies/ncl/index.cfm
a horribly complex study on which i've not had to work much: Sisters Breast Cancer Study
http://www.niehs.nih.gov/research/atniehs/labs/epi/studies/sister/index.cfm

As I've mentioned before, my ideal would be to convince everyone (besides the PIs with whom i've worked most closely ) that having the excellent base of documentation provided by various DDI based solutions (Nesstar, Collectica (which i haven't worked with) is not just a good idea, but necessary.  For many years we used a Cobo l/ Vax based codebook generation which was MUCH better than nothing, although v. long in the tooth.  But that was discarded ~ 14 yrs ago and I regard the replacement (annotated questionnaires, SAS proc contents and various  modes of summarizing each variable) as less than satisfactory.  But the missing values bit is the most frequently used argument against DDI solutions.
Though I'm tempted to take frequencies of just the various missings for each variable and paste them into one of the documentation fields available for each variable.

Bob McConnaughey
Westat/NIEHS Epidemiology Support.

"At times like this, an adult needs a drink."
Dance, Dance, Dance.  H. Murakami



-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.icpsr.umich.edu/pipermail/ddi-users/attachments/20140626/23f04b7c/attachment-0001.html