Re: Riess' prior

From: Peter Nugent (nugent@panisse.lbl.gov)
Date: Thu Feb 27 2003 - 13:31:27 PST

  • Next message: Saul Perlmutter: "Re: info needed ASAP"

    > I'm going to apply this to his dm15 fits. What I still don't understand
    > is just *how* the hell he used it. Here's what I think I know. He's
    > got an mb and an mv; from this he gets an E(B-V). He generates a highly
    > asymmetric error bar on E(B-V) which has a non-gaussian shape by
    > multiplying his measurement error (assumed to be gaussian) by this
    > prior.

    So far so good.

    > What next? Does he *move* E(B-V) to the new value based on the prior?
    > (I.e., this prior is going to generate a probability distribution
    > function whose peak is offset from the old center.) If so, then he's
    > reddening *everything*, which is at odds with his list of AB=0 for most
    > of his supernovae. How does he use it? What did he do?

    As far as I know, from e-mail almost 5 years old, he takes his new E(B-V)
    from the peak value and then computes the second moment of the
    distribution to figure out the size of his uncertainty. Attached is a
    piece of idl code from June of '98 which shows how he does it.

    Good luck,

    Peter

    -- 
    Peter E. Nugent
    Staff Computational Scientist - Scientific Computing Group - NERSC
    Lawrence Berkeley National Laboratory
    M.S. 50F-1650 - 1 Cyclotron Road - Berkeley, CA, 94720-8139
    Phone:(510) 486-6942 - Fax:(510) 486-5812
    E-mail: penugent@LBL.gov - Web: http://supernova.LBL.gov/~nugent
    




    This archive was generated by hypermail 2.1.4 : Thu Feb 27 2003 - 13:31:29 PST