Can do we extend this to distributions with multiple

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: ion • • • Suppose we are given the data X1, X2, ...., Xn and we have already hypothesized the family of distribution from which these observations might come from • • • Possibly from histograms or bar plots Now, we need to estimate the parameters of the hypothesize distribution Two main methods: Method of Moment (MoM): Choose the parameters to match the moments Maximum Likelihood Estimation (MLE): Choose the parameters that maximize the likelihood function 17 Method of Moments (MoM) • • Very general and can be applied in many cases where other estimation methods fail. Example: Suppose we have 5 observations of interarrival times: 3, 1, 4, 3, and 8. Assume that we have hypothesized that these data come from an exponential distribution with some parameter λ • • • Question: How do we determine λ? Note: An exponential distribution with parameter λ has a mean 1/ λ Key Idea: Under MoM, we choose the parameter λ so that the mean (aka first moment) matches with the observed sample mean! • Observed sample mean = (3+1+4+3+8)/5 = 19/5 1 ˆ λ MoM Estimator: = 19 5 ⇔ ˆ λ= 5 19 This method works for any distribution having a single parameter. Can do we extend this to distributions with multiple parameters? 18 General MoM • For k = 1, 2, 3....., the k moment of • Given observations X , ...., X , the the random variable X is defined by: estimated sample k moment is n ￿ ￿ k￿ 1 k mk = (Xi ) µk = E X n i=1 • Suppose the observations X , ...., X are hypothesized to be i.i.d. samples th 1 n th 1 n from a distribution whose p.d.f. f(⋅) has p unknown parameters. • • Step 1: Estimate the first p moments using the data. Let m1, m2, ..., mp denote these estimated sample moments. • • • Note: The estimated moments m1, m2, ..., mp are p numbers Step 2: Analytically compute the first p moments of the hypothesized p.d.f. Let μ1, μ2, ..., μp denote these exact moments. Each μk will be an expression involving the p unknown parameters For many common distributions, the expression for the moments are well-documents. So, you can just look them up! 19 • Step 3: Solve the following system of p equations for the p unknowns parameters: µk = mk k = 1,...
View Full Document

This note was uploaded on 10/26/2010 for the course OR&IE 5580 at Cornell University (Engineering School).

Ask a homework question - tutors are online