Re: observations in different scales

From: Sergey Tarima (stari_at_ms.uky.edu)
Date: 07/10/04

  • Next message: B. D. McCullough: "Re: Fisher vs Neyman&Pearson" 
    Date: Sat, 10 Jul 2004 01:29:46 +0000 (UTC)

    John and Richard:

    You both right that I have to use subject specific information.
    But my original objective was to find THE methods capable to make
    estimation based on independant observations over only ONE random
    variable (in my case this variable is the number events). The problem
    of these observations was that one part of them measured exactly
    (answers from 0 to 96) but the other part know in form of indicator
    functions ("0" or "1+").

    Look at the following manuscript (we submitted it 4 months ago,
    but didn't get response yet; probably we will need to make some
    corrections; I hope minor ones)

    http://www.ms.uky.edu/~stari/TarimaPavlov.pdf

    In section 4 of this manuscript an estimator based on such
    observations was proposed. This estimator is asymptotically unbiased
    with the smallest variance and, moreover, this estimator is a
    well-known Kaplan-Meier one but derived not by means of profile
    likelihood approach but throught the other considerations.

    What I am trying to do now is to provide a comparative analysis of
    this estimator with ... (I do not know which methods should be
    considered).

    The example I suggested is just one of possible applications for the
    methodolgy described in the manuscript.

    I would greatly appreciate any opinions on
    how the comparative analysis should be done.

    With very best regards, Sergey Tarima.

  • Next message: B. D. McCullough: "Re: Fisher vs Neyman&Pearson"