New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

ordinal value specification #862

Open
turbomam opened this Issue Sep 7, 2017 · 14 comments

Comments

Projects
None yet
6 participants
@turbomam
Contributor

turbomam commented Sep 7, 2017

Due to the interest in ordinal value specifications, I'd like to take a small but concrete step as soon as possible. Let's discuss this on Monday September 11th.

create "ordinal value specification" as a subclass of "categorical value specification", http://purl.obolibrary.org/obo/OBI_0001930 with URI http://purl.obolibrary.org/obo/OBI_0002930

proposed definition: "a categorical value specification in which the available categories have a natural order and the distances between the categories is not known" from https://en.wikipedia.org/wiki/Ordinal_data

example of use: "A RNA integrity number of 1 indicates RNA that is more highly degraded than RIN 9"

Future steps:

  • what kind of class would contain multiple ordinal (or categorical) value specifications like 'infant', 'neonate' and 'toddler'? I.e., how to express which ordinal values are legal in a given context?
  • what classes or properties to create, in order to express the rank of a ordinal value specification ('infant' = 2, 'neonate' =1, 'toddler'=3)?
@turbomam

This comment has been minimized.

Show comment
Hide comment
@turbomam

turbomam Sep 11, 2017

Contributor

Gully suggested that an OVS would always be tied to an ordinal measurement datum

Contributor

turbomam commented Sep 11, 2017

Gully suggested that an OVS would always be tied to an ordinal measurement datum

@turbomam

This comment has been minimized.

Show comment
Hide comment
@turbomam

turbomam Sep 11, 2017

Contributor

DEFINITION: where to get definition?
ranking by some natural order sufficient?
distances not known or not of fixed interval... put some of this in editor notes?

Contributor

turbomam commented Sep 11, 2017

DEFINITION: where to get definition?
ranking by some natural order sufficient?
distances not known or not of fixed interval... put some of this in editor notes?

@cstoeckert

This comment has been minimized.

Show comment
Hide comment
@cstoeckert

cstoeckert Sep 11, 2017

Contributor

Discussed on Sept. 11. Want to tie more to data and use example (such RIN value specification) that is clearer. Improve wording for second part.

Contributor

cstoeckert commented Sep 11, 2017

Discussed on Sept. 11. Want to tie more to data and use example (such RIN value specification) that is clearer. Improve wording for second part.

@turbomam

This comment has been minimized.

Show comment
Hide comment
@turbomam

turbomam Sep 11, 2017

Contributor

EXAMPLES OF USE: hardness of minerals... met with limited enthusiasm
indirect immunofluorescence assay for the presence of a particular bacterium: "-", "+", "++", "+++"
probably won't use RIN, as some algorithms appear to generate values on an continuous scale

Contributor

turbomam commented Sep 11, 2017

EXAMPLES OF USE: hardness of minerals... met with limited enthusiasm
indirect immunofluorescence assay for the presence of a particular bacterium: "-", "+", "++", "+++"
probably won't use RIN, as some algorithms appear to generate values on an continuous scale

@Aqua1ung

This comment has been minimized.

Show comment
Hide comment
@Aqua1ung

Aqua1ung Sep 18, 2017

The notion of "physical magnitude" (aka "physical quantity") connotes an ordinal. Always. Results of measurements (of the same magnitude, that is) always constitute an ordered set. [Apologies for the "pronouncement" tone of the posting. I made it in a hurry.]

Aqua1ung commented Sep 18, 2017

The notion of "physical magnitude" (aka "physical quantity") connotes an ordinal. Always. Results of measurements (of the same magnitude, that is) always constitute an ordered set. [Apologies for the "pronouncement" tone of the posting. I made it in a hurry.]

@turbomam

This comment has been minimized.

Show comment
Hide comment
Contributor

turbomam commented Sep 22, 2017

@Public-Health-Bioinformatics

This comment has been minimized.

Show comment
Hide comment
@Public-Health-Bioinformatics

Public-Health-Bioinformatics Sep 25, 2017

I've tried downloading and viewing that file in Protege 5.0.0 on Mac, and protege freezes!? That happening for anyone else?

Public-Health-Bioinformatics commented Sep 25, 2017

I've tried downloading and viewing that file in Protege 5.0.0 on Mac, and protege freezes!? That happening for anyone else?

@zhengj2007

This comment has been minimized.

Show comment
Hide comment
@zhengj2007

zhengj2007 Sep 25, 2017

Contributor
Contributor

zhengj2007 commented Sep 25, 2017

@turbomam

This comment has been minimized.

Show comment
Hide comment
@turbomam

turbomam Sep 26, 2017

Contributor

I created it with Protege 5.2, but I think it should be backwards compatible to whenever Protege started using OWL2. I have opened it in Windows 10 and Ubuntu Linux 16.01, but not Mac... not that the OS is a likely culprit.

Have you been able to open other OWL files in your 5.0 installation recently? (Sorry, annoying question!)

-Mark

Contributor

turbomam commented Sep 26, 2017

I created it with Protege 5.2, but I think it should be backwards compatible to whenever Protege started using OWL2. I have opened it in Windows 10 and Ubuntu Linux 16.01, but not Mac... not that the OS is a likely culprit.

Have you been able to open other OWL files in your 5.0 installation recently? (Sorry, annoying question!)

-Mark

@Public-Health-Bioinformatics

This comment has been minimized.

Show comment
Hide comment
@Public-Health-Bioinformatics

Public-Health-Bioinformatics Sep 27, 2017

Ok then its something odd on my portable. I've opened it in web-protege. Yes, I was opening other files in 5.0 no prob. I'll try rebooting. Thanks for help!
P.s. later was able to open ovs_merged.owl in 5.2 on Mac, though it took some time to load.

Public-Health-Bioinformatics commented Sep 27, 2017

Ok then its something odd on my portable. I've opened it in web-protege. Yes, I was opening other files in 5.0 no prob. I'll try rebooting. Thanks for help!
P.s. later was able to open ovs_merged.owl in 5.2 on Mac, though it took some time to load.

@Public-Health-Bioinformatics

This comment has been minimized.

Show comment
Hide comment
@Public-Health-Bioinformatics

Public-Health-Bioinformatics Sep 27, 2017

A note about traditional ordinal scales like "Beaufort scale" (https://en.wikipedia.org/wiki/Beaufort_scale) that may get mapped over to an interval scale later in their lifetimes. The Beaufort scale has its ordinal buckets translated to a scalar windspeed. Since an ordinal scale has binned its data, only a subset of statistics apply, and so it is important to set these up as ordinal value specifications. (If raw numeric data is available as directly, one can apply a fuller set of stats by avoiding the ordinal scale.)

For a given variable (or survey question) rather than "ord. val. spec. with missing option", I'd side with an "assay attempt" class and "unsuccessful assay" class that would support the given "missing" datum state mentioned in #856. A comment about indicating an item's rank will be there too.

Public-Health-Bioinformatics commented Sep 27, 2017

A note about traditional ordinal scales like "Beaufort scale" (https://en.wikipedia.org/wiki/Beaufort_scale) that may get mapped over to an interval scale later in their lifetimes. The Beaufort scale has its ordinal buckets translated to a scalar windspeed. Since an ordinal scale has binned its data, only a subset of statistics apply, and so it is important to set these up as ordinal value specifications. (If raw numeric data is available as directly, one can apply a fuller set of stats by avoiding the ordinal scale.)

For a given variable (or survey question) rather than "ord. val. spec. with missing option", I'd side with an "assay attempt" class and "unsuccessful assay" class that would support the given "missing" datum state mentioned in #856. A comment about indicating an item's rank will be there too.

@Aqua1ung

This comment has been minimized.

Show comment
Hide comment
@Aqua1ung

Aqua1ung Oct 8, 2017

For the record, and just to reiterate a point that I made in one of the conversations I had this past week: the distinction "categorical value specification"--"ordinal value specification" is untenable imho. Any countable set can be (totally) ordered in many ways, and there is no such thing as a "natural order." Allegiance to a "natural order" betrays, in fact, instinctive/tacit/subliminal allegiance to some ordering standard or other (more or less ephemeral), and while some of these orderings may be more useful (to whom?, and when?) than others, either way they are epistemic artifacts. My suggestion is to refrain from introducing yet another epistemically-motivated distinction.

Aqua1ung commented Oct 8, 2017

For the record, and just to reiterate a point that I made in one of the conversations I had this past week: the distinction "categorical value specification"--"ordinal value specification" is untenable imho. Any countable set can be (totally) ordered in many ways, and there is no such thing as a "natural order." Allegiance to a "natural order" betrays, in fact, instinctive/tacit/subliminal allegiance to some ordering standard or other (more or less ephemeral), and while some of these orderings may be more useful (to whom?, and when?) than others, either way they are epistemic artifacts. My suggestion is to refrain from introducing yet another epistemically-motivated distinction.

@bpeters42

This comment has been minimized.

Show comment
Hide comment
@bpeters42

bpeters42 Oct 9, 2017

bpeters42 commented Oct 9, 2017

@Public-Health-Bioinformatics

This comment has been minimized.

Show comment
Hide comment
@Public-Health-Bioinformatics

Public-Health-Bioinformatics Oct 26, 2017

Agreed, ordinal value specifications are created in our "cave of shadows", and are rather arbitrary buckets of real world dimensional info - unless, like the Beaufort scale, they are further standardized into real number ranges. Statistics have been developed to at least glean something from ordinal representations. Perhaps a reasoner can't do much with ordinal data directly (what is "hot", "medium" or "cold"?), but it could infer the appropriate statistical methods to apply (and in situations where mixed type data is involved).

Public-Health-Bioinformatics commented Oct 26, 2017

Agreed, ordinal value specifications are created in our "cave of shadows", and are rather arbitrary buckets of real world dimensional info - unless, like the Beaufort scale, they are further standardized into real number ranges. Statistics have been developed to at least glean something from ordinal representations. Perhaps a reasoner can't do much with ordinal data directly (what is "hot", "medium" or "cold"?), but it could infer the appropriate statistical methods to apply (and in situations where mixed type data is involved).

@jamesaoverton jamesaoverton changed the title from ordinal value specification: action on Mon Sep 11th? to ordinal value specification Nov 13, 2017

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment