Inception and Development at the University of Virginia

This algorithm has been initiated and developed at the University of Virginia and there are key research articles:

  • Maclarnon (1990)
    The inception of MSER — Marginal Confidence Rule was an initial rule to figure out a truncation point
  • White and Minnox (1994)
    To minimize the half width of the marginal confidence interval about the truncated sample mean
  • Rossetti et al. (1995)
    Proposed an extension of MSER.
  • Spratt (1998)
    Applied the batch mean approach to mitigate serial correlations between output series and introduced MSER-5
  • Franklin et al.(2009)
    Empirical relationship of MSER to the Mean Squared Error (MSE) of a sample mean

Outside UVa

  • Mahajan and Ingalls (2004)
    They evaluated and compared the performances among the five detection methods of a warm-up period and suggested that MSER-5 outperformed others in the most cases.
  • Oh and Park (2006)
    Online detection based on exponential variation rate (EVR) rule was compared with MSER and they concluded MSER performed better
  • Bertoli, Casale, and Serazzi (2007, 2009)
    Developed and implemented Java module of MSER-5.
  • Hoad et al. (2008)
    Up-to-date and the most comprehensive literature review on output analysis of simulation.

Latest Development

Brief introduction

  • Pasupathy and Schmeiser(2010)
    After they showed MSER statistics mathematically confirmed to the MSE criteria in output analysis, they agreed with the general usage of MSER to truncate initial output.
  • Hoad and Robinson (2011)
    First trial of implementing MSER on SIMUL8
  • Hoad et al. (2011)
    Provided a general framework to run automatic output analyzer in the support of MSER-5.
  • Sanchez and White, Jr. (2011)
    After truncating different initial points based on MSER, they showed that relplication/deletion with MSER-truncation using a weighted scheme created a better interval estimation.