a critique of galaxyzoo.org

FIRST THE GOOD

Here are some nice shots I encountered while classifying some galaxies
at galaxyzoo.org:

This one I really liked!

http://setisociety. org/getjpeg8f. jpg

and these were pretty nice too:

http://setisociety. org/getjpeg1. jpg

http://setisociety. org/getjpeg8e. jpg

http://setisociety. org/getjpeg8h. jpg

http://setisociety. org/getjpeg8. jpg

Most of the time they don't look anything as good as the above (and
are incredibly hard to classify) but if you do it (as I still will do)
I highly recommend that you put on some space music and take a cruise
through the universe, you may be the first human observer to do so!

That said

NOW THE BAD:

What's irksome to me is classifying objects that are so at the edge
(if not beyond) of detection. The shots I posted above are easy to
see, but most of the ones offered for examination on the galaxyzoo.org
website are IMO too pixelated/over magnified which must contribute
some degree of possible mis-identification/error. More often than not
I see something that overall looks like an elliptical, but the
pixelation is so high on the fringes that it could be hiding a spiral
arm. Also, I wonder if they could resolve really tight Sa galaxies
over these distances.

I also wonder why there is no consideration of Lenticular galaxies and
how do you distinguish an edge-on lenticular galaxy from an oblate
elliptical or edge-on spiral over these distances!!!!!

Additionally, because of the limited resolution, perhaps one cannot
necessarily determine if an object fits into Hubble's (or others)
classification scheme

http://en.wikipedia.org/wiki/Galaxy_...classification

For example, spheroidal/elliptical galaxies can have varying degreess
of oblateness. The most oblate ones can greatly resemble an edge-on
spiral or lenticular, and with the limited resolution of these shots
(and not knowing if the exposure time is blowing out dust lane
features), I don't think that one can distinguish between the types
over the "billions" of light years they indicate some of these objects
to be at.

Quoting the 1961 Harvard University Press book "Galaxies" by
Harlow Shapley "...For distant galaxies it is difficult to
discriminate between the elongated E7 spheroidal system and some of
the edgewise spirals that show little detail of internal structure and
no granulation into stars, clusters, and nebulosity. In fact, when
arranged in a sequence there is perhaps intrinsically little
difference between these two types, and Hubble has proposed a
connectant form, S0, representing the faintest discernable stage of
spiraling."

My point is that it basically comes down to the galaxy's plane
orientation relative to us, for those that have an edge on appearance,
IMO at these resolutions no distinction can be made regarding whether
it is a spiral or elliptical (AND UNFORTUNATELY NO CONSIDERATION SEEMS
TO HAVE BEEN MADE FOR LENTICULARS AND IRREGULARS!)

Perhaps the best relevant quote from that book is "...English
Astronomer J.H. Reynolds has appropriately emphasized the fact that
practically every galaxy is distinquishable from all others; the
classifications are only convenient shelves, not to be taken too
seriously."

I wonder what this projects scientific value/efficacy will be, based
on the above. One might assume that they would have many people look
at the same object multiple times to get some confidence in the error
for each object. For those objects where it's closer to 50/50 on
reporting (i.e. spiral vs elliptical), those should be thrown in the
error bin, but even when the percentages get closer to 100:0, can one
be sure that there hasn't been a misclassification as an elliptical
when it could be an edge-on spiral (or the un-considered lenticular?)

To summarize, I see that perhaps their greatest project weaknesses
might include the apparent lack of consideration of lenticular types,
no button to click for "irregular galaxies" (that are not interacting
with others), and how one can distinguish the similarity of apperance
of oblate ellipticals, edge-on spirals and lenticulars over these
distances, especially if some of the images they use are extremely
pixelated/over-magnified. Going beyond identifying the chirality of
easily identified spiral galaxies may be an exercise in futility (but
I like the pretty colors and shapes and enjoy going through them too!
:^)

Jason H.

"Jason H." wrote:

FIRST THE GOOD

Here are some nice shots I encountered while classifying some galaxies
at galaxyzoo.org:

This one I really liked!

http://setisociety. org/getjpeg8f. jpg

and these were pretty nice too:

http://setisociety. org/getjpeg1. jpg

http://setisociety. org/getjpeg8e. jpg

http://setisociety. org/getjpeg8h. jpg

http://setisociety. org/getjpeg8. jpg

Most of the time they don't look anything as good as the above (and
are incredibly hard to classify) but if you do it (as I still will do)
I highly recommend that you put on some space music and take a cruise
through the universe, you may be the first human observer to do so!

That said

NOW THE BAD:

What's irksome to me is classifying objects that are so at the edge
(if not beyond) of detection. The shots I posted above are easy to
see, but most of the ones offered for examination on the galaxyzoo.org
website are IMO too pixelated/over magnified which must contribute
some degree of possible mis-identification/error. More often than not
I see something that overall looks like an elliptical, but the
pixelation is so high on the fringes that it could be hiding a spiral
arm. Also, I wonder if they could resolve really tight Sa galaxies
over these distances.

I also wonder why there is no consideration of Lenticular galaxies and
how do you distinguish an edge-on lenticular galaxy from an oblate
elliptical or edge-on spiral over these distances!!!!!

Additionally, because of the limited resolution, perhaps one cannot
necessarily determine if an object fits into Hubble's (or others)
classification scheme

http://en.wikipedia.org/wiki/Galaxy_...classification

For example, spheroidal/elliptical galaxies can have varying degreess
of oblateness. The most oblate ones can greatly resemble an edge-on
spiral or lenticular, and with the limited resolution of these shots
(and not knowing if the exposure time is blowing out dust lane
features), I don't think that one can distinguish between the types
over the "billions" of light years they indicate some of these objects
to be at.

Quoting the 1961 Harvard University Press book "Galaxies" by
Harlow Shapley "...For distant galaxies it is difficult to
discriminate between the elongated E7 spheroidal system and some of
the edgewise spirals that show little detail of internal structure and
no granulation into stars, clusters, and nebulosity. In fact, when
arranged in a sequence there is perhaps intrinsically little
difference between these two types, and Hubble has proposed a
connectant form, S0, representing the faintest discernable stage of
spiraling."

My point is that it basically comes down to the galaxy's plane
orientation relative to us, for those that have an edge on appearance,
IMO at these resolutions no distinction can be made regarding whether
it is a spiral or elliptical (AND UNFORTUNATELY NO CONSIDERATION SEEMS
TO HAVE BEEN MADE FOR LENTICULARS AND IRREGULARS!)

Perhaps the best relevant quote from that book is "...English
Astronomer J.H. Reynolds has appropriately emphasized the fact that
practically every galaxy is distinquishable from all others; the
classifications are only convenient shelves, not to be taken too
seriously."

I wonder what this projects scientific value/efficacy will be, based
on the above. One might assume that they would have many people look
at the same object multiple times to get some confidence in the error
for each object. For those objects where it's closer to 50/50 on
reporting (i.e. spiral vs elliptical), those should be thrown in the
error bin, but even when the percentages get closer to 100:0, can one
be sure that there hasn't been a misclassification as an elliptical
when it could be an edge-on spiral (or the un-considered lenticular?)

To summarize, I see that perhaps their greatest project weaknesses
might include the apparent lack of consideration of lenticular types,
no button to click for "irregular galaxies" (that are not interacting
with others), and how one can distinguish the similarity of apperance
of oblate ellipticals, edge-on spirals and lenticulars over these
distances, especially if some of the images they use are extremely
pixelated/over-magnified. Going beyond identifying the chirality of
easily identified spiral galaxies may be an exercise in futility (but
I like the pretty colors and shapes and enjoy going through them too!
:^)

Jason H.

I like the basic concept but I wonder about the statistical
validity of this process. I assume there are a number of checks and
cross checks. In fact I am sure I noticed selectivity as if testing
participant response last night as I filed through hundreds. Two
images came up twice widely spaced apart. I also wonder about
the omission of irregular types? I have to assume this is covered
or will be at some stage? I like the project and can see people
spending a ton of time with this - its rather interesting. But I like
dull repitive things!

Happy Laughing....

"KLM" wrote in message ...


I wonder what this projects scientific value/efficacy will be, based
on the above. One might assume that they would have many people look
at the same object multiple times to get some confidence in the error
for each object. For those objects where it's closer to 50/50 on
reporting (i.e. spiral vs elliptical), those should be thrown in the
error bin, but even when the percentages get closer to 100:0, can one
be sure that there hasn't been a misclassification as an elliptical
when it could be an edge-on spiral (or the un-considered lenticular?)

To summarize, I see that perhaps their greatest project weaknesses
might include the apparent lack of consideration of lenticular types,
no button to click for "irregular galaxies" (that are not interacting
with others), and how one can distinguish the similarity of apperance
of oblate ellipticals, edge-on spirals and lenticulars over these
distances, especially if some of the images they use are extremely
pixelated/over-magnified. Going beyond identifying the chirality of
easily identified spiral galaxies may be an exercise in futility (but
I like the pretty colors and shapes and enjoy going through them too!
:^)

Jason H.

I like the basic concept but I wonder about the statistical
validity of this process. I assume there are a number of checks and
cross checks. In fact I am sure I noticed selectivity as if testing
participant response last night as I filed through hundreds. Two
images came up twice widely spaced apart. I also wonder about
the omission of irregular types? I have to assume this is covered
or will be at some stage? I like the project and can see people
spending a ton of time with this - its rather interesting. But I like
dull repitive things!


Offering repeated images may be a way of measuring a user's personal
consistency - that way each user's results could be weighted for
reliability.

"KLM" wrote
........
I like the basic concept but I wonder about the statistical
validity of this process..........

I believe that they are only doing a 'first pass' classification catalog.
They will still have a beginning catalog for just about every galaxy down to
25th mag or so, while weeding out the artifacts. I'd bet that any research
use of the catalog will include a second check. However the Journal articles
I've read are full of comments like "we're not sure of the classification of
this galaxy but included it because......".

I assume there are a number of checks and
cross checks. In fact I am sure I noticed selectivity as if testing
participant response last night as I filed through hundreds. Two
images came up twice widely spaced apart......

Somewhere in the documentation it says that *each* candidate will be typed
by several people. If there is a lack of consensus one of the pro's will
make the final call. Also, remember that the candidates were selected by
software, and the program may well give two ID's to the same non-stellar
fuzzy spot, especially if there are two or more data runs covering the same
area of the sky.

I also wonder about
the omission of irregular types? I have to assume this is covered
or will be at some stage?

They ask that you e-mail them the ID number of any galaxy that you believe
does not fall into the offered categories, or that is very unusual, and they
will review it for a final determination. So far I've found one polar ring
galaxy, and another with a full outer ring; I e-mailed them about these.
Also one 'galaxy' I was offered was obviously a dim comet - and I sent yet
another e-mail. I assume that this process will cover the irregulars. I
assume that the they are classifying ellipticals and lenticulars as
basically the same type of galaxy, i.e., one lacking in gas, dust, or recent
star formation. I also assume that they know that the smaller and dimmer the
candidate, the more likelihood of classification error. Just weeding out the
real galaxies from the artifacts is of value.

George Normandin

They ask that you e-mail them the ID number of any galaxy that you believe
does not fall into the offered categories, or that is very unusual, and they
will review it for a final determination. So far I've found one polar ring
galaxy, and another with a full outer ring; I e-mailed them about these.

I've found one polar ring and three outer ring galaxies. I *did not*
email them in. You guys should see the one I did call in. (wooooo)

I'm going with these precepts:
(a) If it's irregular or lenticular call it elliptical.
(b) If the eccentricity is greater than 0.9 call it a edge-on.
(I've had to get out the Vernier caliper
a couple of times to measure.)
(c) If there's absorption material at all near the nucleus call it
a spiral and make a stab at the chirality.
(d) Don't call them merging unless there is some deformation
in one the galaxies.

Ben

"Ben" wrote
...
I've found one polar ring and three outer ring galaxies. I *did not*
email them in. You guys should see the one I did call in. (wooooo)

Cool Ben!! I'm sure there's lots more in all this SDSS data.


I'm going with these precepts........

Sounds like it will work to me!

I've gotten an entire class of high school advance science kids turned on to
'the zoo' and thier going nuts classifying galaxies!

George N

I've gotten an entire class of high school advance science kids turned on to
'the zoo' and thier going nuts classifying galaxies!

George N

Good idea George - need all the help we can get. I wonder what
they will do with the website when and if we ever reach the terminus.

Do you suppose they will run select examples again in order to refine
further?

Ben

Ben wrote:
I'm going with these precepts:
(a) If it's irregular or lenticular call it elliptical.
(b) If the eccentricity is greater than 0.9 call it a edge-on.
(I've had to get out the Vernier caliper
a couple of times to measure.)

nitpick
Do you really mean eccentricity? An ellipse with eccentricity 0.9 is
barely twice as long as it is wide--what on the tuning fork I would say
is E5 or so. Diagrams seem to vary on that.

I'm guessing you mean something like the galaxy is 10-plus times longer
than it is wide. If so, I'm picking a really minor nit (since I think
most people will guess what you mean).
/nitpick

--
Brian Tung
The Astronomy Corner at http://astro.isi.edu/
Unofficial C5+ Home Page at http://astro.isi.edu/c5plus/
The PleiadAtlas Home Page at http://astro.isi.edu/pleiadatlas/
My Own Personal FAQ (SAA) at http://astro.isi.edu/reference/faq.html

nitpick
Do you really mean eccentricity? An ellipse with eccentricity 0.9 is
barely twice as long as it is wide--what on the tuning fork I would say
is E5 or so. Diagrams seem to vary on that.

Yeah, Brian the eccentricity provides and index of the tilt.

acos 0.9 = 25.84d and that's "edge-on" enough for me. I don't know
about the people at galaxyzoo.


I'm guessing you mean something like the galaxy is 10-plus times longer
than it is wide. If so, I'm picking a really minor nit (since I think
most people will guess what you mean).

What I've been doing is taking a Vernier caliper and measuring both
axes,
divide both by 2 to get a and b, then

e = (a^2 - b^2)^0.5 / a

This only goes down when the eccentricity is really close to 0.9.
If it doesn't make the cut I call it a spiral and look at the fringes
of the object to try and make a guess at any incipient rotation.

Ben

"Brian Tung" wrote in message
...
: Ben wrote:
: I'm going with these precepts:
: (a) If it's irregular or lenticular call it elliptical.
: (b) If the eccentricity is greater than 0.9 call it a edge-on.
: (I've had to get out the Vernier caliper
: a couple of times to measure.)
:
: nitpick
: Do you really mean eccentricity? An ellipse with eccentricity 0.9 is
: barely twice as long as it is wide--what on the tuning fork I would say
: is E5 or so. Diagrams seem to vary on that.
:
: I'm guessing you mean something like the galaxy is 10-plus times longer
: than it is wide. If so, I'm picking a really minor nit (since I think
: most people will guess what you mean).
: /nitpick

add to nitpick
ecc = sqrt(1 - minor^2 /major^2)
minor axis = 0.436/major axis

What I'm guessing he's trying to say is a circular *orbit* seen roughly
edge-on has an *apparent* eccentricity of 0.9, leading to the strange
conclusion that all orbits are circular.
When applied to a galaxy there may be some truth in that, I've
never see an image of a galaxy that was truly eccentric but many
that are apparently so, the assumption always being that the galaxy is
circular. Where the galaxy is lenticular we've added a third dimension
onto the projection and so a circular galaxy seen edge-on has an
apparent "thickness", which should not be confused with eccentricity.

http://www.germes-online.com/direct/...ner_Set.jpgOne plate is both lenticular and eccentric, all are apparently eccentric./add to nitpick