Semantic Web Health Care and Life Sciences Interest Group Teleconference

05 Aug 2015

See also: IRC log


Paul_Courtney, EricP, Lloyd, DBooth
David Booth


<trackbot> Date: 05 August 2015

<ericP> hey gang

hi eric


Decimal precision

Eric: Not ambiguity, but an invitation for systems to do the wrong thing. Not that 12.0000 is ambiguous, but that it's easy for someone to say that and not mean it. If you call out the precision in a separate property they you eliminate the chances that someone will printf 4.2 , and you make it directly queryable. But the downside is that digits are a coarse way of providing error bars, instead of +0.3 -0.1. Once we find the use cases we can ask if we wa

nt to meet them and how.

lloyd: If someone spits out precision that isn't there, they may do that with an explicit attribute. If they don't declare precision and we discard trailing zeros. Implementations that care may have to look for it both explicitly and implicitly.
... I would expext most systems to display what they are given.

david: I would expect data to have fields of standard widths.

paul: I'm looking at my latest medical record, and my height is shown to three digits.

lloyd: i don't see the value in having explicit precision.

david: what use cases for knowing precision?

lloyd: primarily comparing values. Two values 3.0 and 3.001 are close enough.

david: But we routinely compare for 'close enough' anyway, such as 'within 1%'

eric: Given that medicine already has conventions around the use of trailing zeros, we wouldn't be done with a precision bit is added . . . [ eric dropped]

<ericP> !

eric, coming back?

eric: it would have to go back to make the lab equipment emit that information.

lloyd: Possible that there may be 1, 2 or 10 extraneous digits that are not significant. But it typically gets corrected downstream, or viewed as an error.

paul: if the digits of precision are significant, then what's the harm in making them explicit?

lloyd: one FHIR principle, is that FHIR does not drive system behavior. FHIR won't make people do things that the
... that they are not doing already.
... existing systems don't explicitly capture precision, so we can't make a good argument of requiring it.

david: what if interoperability conflicts with existing system behavior?

lloyd: FHIR helps systems exchange whatever they have. Conformance framework allows the bar to be raised for better interop.
... FHIR has a very low bar to entry, then makes it easy to make incremental improvements.

david: I agree that an big bang assumption of achieving full interop would be too much, but I think it's important to have a bias toward encouraging interop.
... it seems a shame to have to represent numbers as strings. what we'd really like is a precisionDecimal datatype that is widely implemented. But we don't have that.
... In any case, I don't think this will impact the FHIR RDF decision that we made yesterday.


Summary of Action Items

[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.140 (CVS log)
$Date: 2015/08/05 16:06:44 $

Scribe.perl diagnostic output

[Delete this section before finalizing the minutes.]
This is scribe.perl Revision: 1.140  of Date: 2014-11-06 18:16:30  
Check for newer version at http://dev.w3.org/cvsweb/~checkout~/2002/scribe/

Guessing input format: RRSAgent_Text_Format (score 1.00)

No ScribeNick specified.  Guessing ScribeNick: dbooth_
Inferring Scribes: dbooth_
Present: Paul_Courtney EricP Lloyd DBooth
Found Date: 05 Aug 2015
Guessing minutes URL: http://www.w3.org/2015/08/05-hcls-minutes.html
People with action items: 

[End of scribe.perl diagnostic output]