Dismiss Notice
Join Physics Forums Today!
The friendliest, high quality science and math community on the planet! Everyone who loves science is here!

Confidence levels.

  1. Mar 8, 2004 #1

    wolram

    User Avatar
    Gold Member

    http://arxiv.org/abs/astro-ph/0401198

    How many cosmological parameters?
    Andrew R. Liddle
    Astronomy Centre, University of Sussex, Brighton BN1 9QH, United Kingdom
    29 January 2004
    ABSTRACT
    Constraints on cosmological parameters depend on the set of parameters chosen to define the model which is compared with observational data. I use the Akaike and Bayesian information criteria to carry out cosmological model selection, in order to
    determine the parameter set providing the preferred fit to the data. Applying the information criteria to the current cosmological data sets indicates, for example, that spatially-flat models are statistically preferred to closed models, and that possible
    running of the spectral index has lower significance than inferred from its confidence.
    ---------------------------------------------------------------------
    this is an interesting paper, I'm not sure if the A and B information
    criteria are better than others, but this paper questions the
    accuracy of CLs "confidence levels", does anyone have an opinion
    on LIDDLEs paper?
     
  2. jcsd
  3. Mar 9, 2004 #2

    Nereid

    User Avatar
    Staff Emeritus
    Science Advisor
    Gold Member

    Nice paper! Very clear and readable.

    I particularly liked the discussion of publication bias - we here at PF (A&C) have seen a more extreme version of this, in that a) many researchers like to 'blow their trumpet' and the journos they talk with like to write catchy articles, and b) PF members will pick and write posts about only the most arresting of these journos' articles.

    Of course, Liddle is simply saying what I've been saying :wink: (though he says it with the backing of solid results from information theory, and writes a darn sight better than I can )
     
  4. Mar 16, 2004 #3

    Nereid

    User Avatar
    Staff Emeritus
    Science Advisor
    Gold Member

    Today's the 16th, and there have been 62 views of this thread, which is far more than the ~10 per post of a 'normal' thread.

    What do all you other readers (PF members) think about the paper which wolfram so kindly posted here for us? Is it really cool, or is it not?? [?]
     
  5. Mar 19, 2004 #4

    wolram

    User Avatar
    Gold Member

    by NEREID.

    I particularly liked the discussion of publication bias - we here at PF (A&C) have seen a more extreme version of this, in that a) many researchers like to 'blow their trumpet' and the journos they talk with like to write catchy articles, and b) PF members will pick and write posts about only the most arresting of these journos' articles.
    ---------------------------------------------------------------------
    it is probibily worthy of a book on "how to avoid catchy science",
    for some it is easy to spot media seeking researchers, but others
    myelf included can fall into their influence, best advice wait
    for reaction to publication, ask PF for opinion.
     
  6. Mar 19, 2004 #5

    selfAdjoint

    User Avatar
    Staff Emeritus
    Gold Member
    Dearly Missed

    Nereid, it is truly cool. I particularly liked the part about publication bias, which certainly should be taken to heart! And I wonder about applying the same information criteria (AIC and BIC) to other areas with a lot of data. For example cognitive performance data and the g model.
     
Know someone interested in this topic? Share this thread via Reddit, Google+, Twitter, or Facebook

Have something to add?