+ Reply to Thread
Results 1 to 8 of 8

Thread: New to Bayesian Stats

  1. #1
    Points: 375, Level: 7
    Level completed: 50%, Points required for next Level: 25

    Posts
    4
    Thanks
    0
    Thanked 0 Times in 0 Posts

    Talking New to Bayesian Stats




    Hello! I'm fairly new to Bayesian stats and will be using WinBugs to analyze spatial data of West Nile virus. I have a lot to learn and was hoping this forum could answer some of my questions. Thanks all!

    Josiah

  2. #2
    Beep
    Points: 58,109, Level: 100
    Level completed: 0%, Points required for next Level: 0
    Awards:
    Discussion EnderPosting AwardCommunity AwardMaster TaggerFrequent Poster
    Dason's Avatar
    Location
    Ames, IA
    Posts
    10,683
    Thanks
    249
    Thanked 2,054 Times in 1,754 Posts

    Re: New to Bayesian Stats

    Hi!



    The question you asked was interesting - I hope you know you're in for some pretty complex stuff if you want to do your analysis right! Bayesian statistics is pretty awesome though.
    001100010010011110100001101101110011

  3. #3
    Point Mass at Zero
    Points: 6,886, Level: 54
    Level completed: 68%, Points required for next Level: 64
    ledzep's Avatar
    Location
    Berks,UK
    Posts
    648
    Thanks
    174
    Thanked 132 Times in 130 Posts

    Re: New to Bayesian Stats

    Hope we'll get chance to learn more about White Nile virus in the due course. Watch out for your hyper-parameters though! Sometimes uninformative priors for you hyper parameters in fact turns out to be informative!!! Welcome to the club.
    Oh Thou Perelman! Poincare's was for you and Riemann's is for me.

  4. #4
    Beep
    Points: 58,109, Level: 100
    Level completed: 0%, Points required for next Level: 0
    Awards:
    Discussion EnderPosting AwardCommunity AwardMaster TaggerFrequent Poster
    Dason's Avatar
    Location
    Ames, IA
    Posts
    10,683
    Thanks
    249
    Thanked 2,054 Times in 1,754 Posts

    Re: New to Bayesian Stats

    Quote Originally Posted by ledzep View Post
    Sometimes uninformative priors for you hyper parameters in fact turns out to be informative!!! Welcome to the club.
    Uninformative priors are pretty much always informative in one way or another.
    001100010010011110100001101101110011

  5. #5
    Point Mass at Zero
    Points: 6,886, Level: 54
    Level completed: 68%, Points required for next Level: 64
    ledzep's Avatar
    Location
    Berks,UK
    Posts
    648
    Thanks
    174
    Thanked 132 Times in 130 Posts

    Re: New to Bayesian Stats

    Quote Originally Posted by Dason View Post
    Uninformative priors are pretty much always informative in one way or another.
    May be they should be called something like "informative uninformative priors".
    Oh Thou Perelman! Poincare's was for you and Riemann's is for me.

  6. #6
    Beep
    Points: 58,109, Level: 100
    Level completed: 0%, Points required for next Level: 0
    Awards:
    Discussion EnderPosting AwardCommunity AwardMaster TaggerFrequent Poster
    Dason's Avatar
    Location
    Ames, IA
    Posts
    10,683
    Thanks
    249
    Thanked 2,054 Times in 1,754 Posts

    Re: New to Bayesian Stats

    I kind of like the term 'reference' priors.
    001100010010011110100001101101110011

  7. #7
    R must die
    Points: 22,299, Level: 92
    Level completed: 95%, Points required for next Level: 51
    noetsi's Avatar
    Posts
    4,150
    Thanks
    245
    Thanked 653 Times in 630 Posts

    Re: New to Bayesian Stats

    Quote Originally Posted by Dason View Post
    Uninformative priors are pretty much always informative in one way or another.
    Isn't that an oxymoron....

    It reminds me of John Fox's advice that outliers are nearly always useful, because they clue you into to problems in your theory or misspecified model you really need to know.
    Hail A.I.M!

  8. #8
    Beep
    Points: 58,109, Level: 100
    Level completed: 0%, Points required for next Level: 0
    Awards:
    Discussion EnderPosting AwardCommunity AwardMaster TaggerFrequent Poster
    Dason's Avatar
    Location
    Ames, IA
    Posts
    10,683
    Thanks
    249
    Thanked 2,054 Times in 1,754 Posts

    Re: New to Bayesian Stats


    We call some priors uninformative because they don't influence the posterior distribution much or they don't give us much information about the parameter of interest beforehand. But it's pretty much impossible to eliminate all information - something that is "uniformative" about a parameter in one sense of the word might not be uninformative about a transformation of that parameter. I'm not a huge fan of the 'uninformative' prior terminology but it does at least get across sort of what one is trying to do with that prior - not provide much information about the parameter before hand.
    001100010010011110100001101101110011

+ Reply to Thread

           




Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts






Advertise on Talk Stats