Galaxy Zoo Talk

PhotoZ Error Bar

  • ChrisMolloy by ChrisMolloy

    This is AGZ00047hl, which is the target galaxy in the image.

    enter image description here

    AGZ00047hl has a PhotoZ of z=0.210344 +/-0.01863.

    enter image description here

    The object at the top edge, with a slight overlap of AGZ00047hl is SDSS J012011.71-205354.8. This has a PhotoZ of z=0.119722 +/--1000.

    enter image description here

    What does the error bar of +/--1000 mean for SDSS J012011.71-205354.8. The error bar only decreases?

    Posted

  • Capella05 by Capella05 moderator in response to ChrisMolloy's comment.

    Hi Chris,

    I believe it means that the photoZ for the 2nd object is unusable.

    There appears to be a lot of confusion about photoZ going around, and I have been meaning to write a glossary entry about it. PhotoZ is an estimate calculated by an algorithm, using the colour of other objects / galaxies in the vicinity. It needs several variables in order to make the calculation, so there is a error margin.

    Posted

  • Budgieye by Budgieye moderator

    Here is what SDSS says about PhotoZ. Warning: lots of big words!

    http://skyserver.sdss.org/dr7/en/help/docs/algorithm.asp?key=photoz

    To me, it looks like the photoz error of -1000 looks like a math operation error.

    With the huge error margins in photoz, I think anyone who is not colour blind can do an equally good job. I look at the target galaxy and think, " fuzzy detail so at least z=0.1, turning orange so z=0.2, but too too orange, so not 0.3. With the smaller galaxy, I think "Well, it looks about the same"

    The accurate way of estimating the distance of a galaxy is by doing a spectral chart, which is time-consuming and costs money. Only a million and half spectra have been done !!! but there are a billion objects to be looked at. http://www.sdss3.org/dr9/
    So no spectral data is available for the less interesting galaxies.

    But if someone is going to do a computer search of yellow-orange galaxies, they can't put "yellow-orange" in the search parameters, they need a number. So for a computer

    yellow-orange is photoz 0.1-0.3

    galaxy is "a source which is not round"

    Unfortunately, it might be mean two orange stars close together, or a extremely distant orange quasar near something else.

    Posted

  • ChrisMolloy by ChrisMolloy

    Thanks Capella05, Budgieye,

    With the huge error margins in photoz, I think anyone who is not colour blind can do an equally good job. I look at the target galaxy and think, " fuzzy detail so at least z=0.1, turning orange so z=0.2, but too too orange, so not 0.3. With the smaller galaxy, I think "Well, it looks about the same"

    Very informative Budgieye. Very similar to what you'd posted in the GZ forums on the Colours of Galaxies in SDSS: Redshift Chart. I'd completely forgotten your post. Will use it as a cross reference/sanity check next time for PhotoZ.

    Posted

  • Budgieye by Budgieye moderator

    Thank you !! 😃 😃

    updated version of the chart is

    Astrophysics for Galaxy Zoo Talk - redshift z PhotoZ spectra emissions AGN

    Galaxy Redshift Chart http://talk.galaxyzoo.org/#/boards/BGZ0000001/discussions/DGZ0000ulp?page=2

    The links work better than in the forum version.

    I'll put this reply into it.

    Posted