1. I've been setting the threshold incorrectly for testing the likelihoods. I need to do the following. This should actually speed things up a bunch because I'll need to calculate the likelihoods on a lot fewer objects.

-----a) Pick a set of targets (Threshold Set) in a known area from the test objects, say -1.25 < dec < 1.25 and 1 degree in ra spaced across the stripe (-20, -10, 0, 10, 20). The Threshold Set of targets occupies 5*2.5 = 12.5 deg

^{2}of area.

-----b) Calculate the (new) likelihoods on the Threshold Set.

-----c) Determine a threshold (TH) of (X/deg

^{2}) by picking the objects with the top Y = X*12.5 likelihoods in the Threshold Set. The threshold will be the min(likelihood_ratio) of these Y objects.

-----d) Calculate the (new) likelihoods on all objects in the 3pc catalog that are on Stripe-82 (Truth Table Set).

-----e) We targeted at 80 targets/deg

^{2}in this region, so we can get the effective area (A) of the Truth Table Set by dividing the # of objects (Z) / 80: A (deg

^{2}) of Truth Table Set = Z/(80 targets/deg

^{2})

-----f) Determine the (new) targeted quasars (TQ) by saying that any quasar in the Truth Table Set with likelihood greater than TH.

-----g) The QSOs /deg

^{2}= TQ / A.

2. Joe fixed his luminosity function code, so I can now run it to generate the Hopkins, Richards, Hernquist (2007) luminosity function.

3. Joe thinks we don't need to put a brightness cut on the quasar inputs. Gonna see if that changes things

4. David gave me SDSS DR7 Quasars to use as inputs into the Monte Carlo.

5. I should look at something in the target files called x2_star. This is the chi-squared distance of the objects from the stellar locus, and probably all likelihood targets have a x2_star which is big. We might be able to reduce the likelihood calculations by only calculating likelihoods on objects which pass a certain x2_star threshold.

## No comments:

## Post a Comment