Variance Study - PowerPoint PPT Presentation

1 / 7
About This Presentation
Title:

Variance Study

Description:

n aG(1/b 1) a, b, n. Weibull. No allows 0. s2. m. m. ... Erlang (Gamma w/ b = k) Yes. 1/(bq2) 1/q. b, q. Gamma. No. l. l. l. Exponential. No (max-min)2/2 ... – PowerPoint PPT presentation

Number of Views:23
Avg rating:3.0/5.0
Slides: 8
Provided by: mpeter2
Category:
Tags: study | variance

less

Transcript and Presenter's Notes

Title: Variance Study


1
Variance Study
  • M. Peter Jurkat
  • CS452/Mgt532 Simulation for Managerial Decisions
  • The Robert O. Anderson Schools of Management
  • University of New Mexico

2
Requirements
  • Queuing theory says queues get long when
    utilization rises above .6-.7 based on average
    inter-arrival time and average service time
  • No account of inter-arrival time variance
  • To study effect of variance need to separate
    effects of inter-arrival mean from inter-arrival
    variance gt seek inter-arrival distribution for
    which variance can be changed w/o changing the
    mean

3
Examine Suitability of Distributions for Variance
Study
4
Strategy
  • Queuing theory does not have formulas for
    performance measures (r, L, wq, ) in terms of
    inter-arrival variance gt need to simulate
  • Banks4Example2-1SSQ.gp is the simples model
    change distribution in x-act GENERATE block to
    Gamma Distribution
  • My version of GPSS World does not compile with a
    Gamma distribution gt use Excel to generate Gamma
    Distributions and enter into GPSS World as an
    empirical distribution

5
Reconcile Parameters
  • Banks et al uses a different Gamma Distribution
    formulation that shows independence of variance
    from mean formulation in GPSS World and Excel
    does not
  • GPSS World and Excel use the same formulation
    with different notation, but Excel does not show
    formulas for mean and variance
  • Need to reconcile parameters in all three

6
Parameter Comparison
7
Excel Parameters in Terms of Distribution Mean
and Variance
  • From argument of G() function aE aG bB
  • From exponent of e we get bBqB 1/bE
  • The mean, m, and variance, s2, are the same
    quantity regardless of the formulation
  • From the Banks et al formulation m 1/qB and s2
    1/(bBqB2) 1/((bBqB) qB) 1/((1/bE)(1/m))
    bEm gt bE s2/m
  • Also s2 1/(bBqB2) gt qB2 1/(bB s2) 1/m2 so
    bB s2 m2 and bB m2/ s2 aE
  • These values of aE and bE in terms of m and s can
    now be used in Excel to develop Gamma
    Distributions whose means are independent of
    their variances see GammaDistributions.xls
Write a Comment
User Comments (0)
About PowerShow.com