Data sets for proper motions in the Galactic planebcsv

In the context of Nina getting the best data sets for doing proper motion in the Galactic plane,
various discussions (Nina, David, Zeljko) have occurred on what data sets to combine with PS1
data for the most productive proper motion constraints. It seems there should be two sets of criteria:

-- what does the data set add in terms of  dmu ~ FWHM / (S/N) / Depoch

-- how painful are systematic issues, such as photo-plate boundaries etc..


SDSS-overlap areas
  -- take SDSS as an additional epoch
      * SDSS positions are based on the r-band; check what information arises
          from the g band position measurements
  -- take mu(SDSS-POSS) as a prior, rather than deal with (a,d)(POSS) itself
       * work out how to deal with the fact that mu(SDSS-POSS) may have systematics
UKIDSS: could be very interesting the galactic plane:
Questions:
  -- when are the observation epochs of the survey
  --  practical access: get the whole darn thing and add to LSD?
  -- some links
       http://www.astrogrid.org/agpython.html
       http://surveys.roe.ac.uk/wsa/theSurveys.html

Areas that have neither UKIDSS nor SDSS?
   -- take 2MASS, or is it not worth it?
   --

RR Lyrae in PS1

Nina's working out lots of things about how to do QSO variability right.

Questions arises what would need to be changed to apply same formalism to RR Lyrae.

If one uses Gaussian process (Hogg's advice), then one needs to specify the appropriate
covariance function.

Hogg recommends the covariance corresponding to a damped oscillator.

HWR's thoughts:

Take a covariance that corresponds to range of periods of near-sinusoidal oscillation.
It will look periodic for Dt's corresponding to a few periods, but look like
white noise expectations for long time-intervals.
That way, for short time intervals it matches well, but one is not stuck with the cumbersome
period matching over very long periods.

For thinking about how to write down a suitable GP covariance function, I found it useful
to do a lot of staring at the plots from Sesar's 2010 paper (his Fig 9)

It nicely shows:

• the level of lightcurve asymmetry: slow fall, very rapid rise
• the fact that the g-r color changes drastically but the i-z not at all (just Teff?)

Project thoughts

1) Kinematic Consensus:
 

Basic idea: seach for stream members in Integral-of-Motion space, by making the
  grossly simplified assumption of spherical potential and some plausible radial profile.
  The first issue to be explored is: we have excellent (alph,del), good D_photo, v_los, but no v_perp.
  How does that mapping back into integral of motion space look like?

  Hogg will see how far he gets; Zhitai will start perhaps in this direction [TBD]

  Document at https://github.com/davidwhogg/KinematicConsensus/tree/master/documents


2) Long (and for me fruitful) discussion with Bovy led to rediscovery of math in Helmi & White 99
    Question remains: what is the best practical way to "fit streams" (and infer potential)
    from incomplete, or at least imperfect, (x,v) information on relatively few stars.
    In the end, it all boils down to being able to write down L( (x,v), J,theta,Dt's) where the Dt's
    are the times at which the stars were lost.  [Bovy's blackboard notes]

3) Astrometric Übercalibration thoughts:

notes from meeting with Oxonians