different MLE gives different answers????

Discussion in 'Scientific Statistics Math' started by maria.hartberg, Apr 23, 2008.

  1. 1. What is the difference between goodness of fit and maximum

    2. If I have understood it correctly in logistic regression one can
    use different maximum likelihood equations in order to find the best
    fit. So my questions is wouldn’t the answer differ depending on what
    maximum likelihood equation one uses?

    Best wishes Maria
    maria.hartberg, Apr 23, 2008
    1. Advertisements

  2. Maximum likelihood is one of the methods that is
    used to measure fit, and to compare fits. Sometimes
    that is called a "goodness of fit" test. When the comparison
    is made to null-hypotheses such as "coefficient equals zero",
    the result not usually called a goodness-of-fit test.
    Huh? Logistic regression has one standard form.
    Different MLE equations .... Would you mean,
    for different variables? for different ways of adjusting
    for zero-cells, if you are following a discrete solution?

    The latter is more common in log-linear modeling, which
    is a cousin of logistic regression.
    I think you are under some mis-impression about the
    existence of "different maximum likelihood equations"
    or their nature. Would you want to be more specific,
    or have I said enough?
    Richard Ulrich, Apr 24, 2008
    1. Advertisements

  3. maria.hartberg

    Paul Rubin Guest

    Could mean different algorithms for finding the MLE estimates (?). Or
    maybe different methods of initializing the algorithm?

    Maybe this refers to different link functions?

    Paul Rubin, Apr 24, 2008
  4. maria.hartberg

    Herman Rubin Guest

    Your first question suggests that you study your
    material again.

    As to the second question, maximum likelihood is maximum
    likelihood. For the same model, there is almost always one
    absolute maximum, but a computational procedure can
    converge to a relative maximum which gives a value below
    the absolute maximum.
    Herman Rubin, Apr 25, 2008
    1. Advertisements

Ask a Question

Want to reply to this thread or ask your own question?

You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.