ERGM conditional form Much easier to calculate delta (change statistics)

Slides:



Advertisements
Similar presentations
Introduction to Monte Carlo Markov chain (MCMC) methods
Advertisements

Contrastive Divergence Learning
Modeling of Data. Basic Bayes theorem Bayes theorem relates the conditional probabilities of two events A, and B: A might be a hypothesis and B might.
The Maximum Likelihood Method
Sampling: Final and Initial Sample Size Determination
Ch 11. Sampling Models Pattern Recognition and Machine Learning, C. M. Bishop, Summarized by I.-H. Lee Biointelligence Laboratory, Seoul National.
N.D.GagunashviliUniversity of Akureyri, Iceland Pearson´s χ 2 Test Modifications for Comparison of Unweighted and Weighted Histograms and Two Weighted.
Bayesian statistics – MCMC techniques
A gentle introduction to Gaussian distribution. Review Random variable Coin flip experiment X = 0X = 1 X: Random variable.
Course overview Tuesday lecture –Those not presenting turn in short review of a paper using the method being discussed Thursday computer lab –Turn in short.
Parametric Inference.
Today Today: Chapter 9 Assignment: 9.2, 9.4, 9.42 (Geo(p)=“geometric distribution”), 9-R9(a,b) Recommended Questions: 9.1, 9.8, 9.20, 9.23, 9.25.
Standard error of estimate & Confidence interval.
Review of Statistical Inference Prepared by Vera Tabakova, East Carolina University.
Model Inference and Averaging
Random Sampling, Point Estimation and Maximum Likelihood.
Bayesian inference review Objective –estimate unknown parameter  based on observations y. Result is given by probability distribution. Bayesian inference.
Machine Learning Lecture 23: Statistical Estimation with Sampling Iain Murray’s MLSS lecture on videolectures.net:
CSC321: 2011 Introduction to Neural Networks and Machine Learning Lecture 11: Bayesian learning continued Geoffrey Hinton.
Lab 3b: Distribution of the mean
A Comparison of Two MCMC Algorithms for Hierarchical Mixture Models Russell Almond Florida State University College of Education Educational Psychology.
Lecture 2 Review Probabilities Probability Distributions Normal probability distributions Sampling distributions and estimation.
Molecular Systematics
Lecture PowerPoint Slides Basic Practice of Statistics 7 th Edition.
MCMC in practice Start collecting samples after the Markov chain has “mixed”. How do you know if a chain has mixed or not? In general, you can never “proof”
MCMC reconstruction of the 2 HE cascade events Dmitry Chirkin, UW Madison.
Week 41 How to find estimators? There are two main methods for finding estimators: 1) Method of moments. 2) The method of Maximum likelihood. Sometimes.
Point Estimation of Parameters and Sampling Distributions Outlines:  Sampling Distributions and the central limit theorem  Point estimation  Methods.
Logistic Regression Saed Sayad 1www.ismartsoft.com.
1 Chapter 8: Model Inference and Averaging Presented by Hui Fang.
10.1 – Estimating with Confidence. Recall: The Law of Large Numbers says the sample mean from a large SRS will be close to the unknown population mean.
R. Kass/W03 P416 Lecture 5 l Suppose we are trying to measure the true value of some quantity (x T ). u We make repeated measurements of this quantity.
Bayesian II Spring Major Issues in Phylogenetic BI Have we reached convergence? If so, do we have a large enough sample of the posterior?
SIR method continued. SIR: sample-importance resampling Find maximum likelihood (best likelihood × prior), Y Randomly sample pairs of r and N 1973 For.
Generalization Performance of Exchange Monte Carlo Method for Normal Mixture Models Kenji Nagata, Sumio Watanabe Tokyo Institute of Technology.
Markov Chain Monte Carlo in R
Inference: Conclusion with Confidence
The Maximum Likelihood Method
Oliver Schulte Machine Learning 726
MCMC Output & Metropolis-Hastings Algorithm Part I
Optimization of Monte Carlo Integration
Probability Theory and Parameter Estimation I
Advanced Statistical Computing Fall 2016
Inference: Conclusion with Confidence
7-1 Introduction The field of statistical inference consists of those methods used to make decisions or to draw conclusions about a population. These.
Model Inference and Averaging
Introduction to the bayes Prefix in Stata 15
The Maximum Likelihood Method
Maximum Likelihood Estimation
CS 2750: Machine Learning Expectation Maximization
Latent Variables, Mixture Models and EM
Bayesian inference Presented by Amir Hadadi
Markov chain monte carlo
The Maximum Likelihood Method
Estimation Maximum Likelihood Estimates Industrial Engineering
Remember that our objective is for some density f(y|) for observations where y and  are vectors of data and parameters,  being sampled from a prior.
CONCEPTS OF ESTIMATION
Introduction to Inference
More about Posterior Distributions
Lecture 2 – Monte Carlo method in finance
EC 331 The Theory of and applications of Maximum Likelihood Method
Where did we stop? The Bayes decision rule guarantees an optimal classification… … But it requires the knowledge of P(ci|x) (or p(x|ci) and P(ci)) We.
Ch13 Empirical Methods.
EM for Inference in MV Data
EM for Inference in MV Data
Markov Networks.
Maximum Likelihood We have studied the OLS estimator. It only applies under certain assumptions In particular,  ~ N(0, 2 ) But what if the sampling distribution.
Statistical Model A statistical model for some data is a set of distributions, one of which corresponds to the true unknown distribution that produced.
Statistical Model A statistical model for some data is a set of distributions, one of which corresponds to the true unknown distribution that produced.
Presumptions Subgroups (samples) of data are formed.
Presentation transcript:

ERGM conditional form Much easier to calculate delta (change statistics)

Metropolis algorithm Generates a random sample from a given distribution (any distribution!) Given a current state of the network, randomly propose a new state, then randomly accept the change or stay put. Probability of acceptance is: If proposal is a single edge toggle, then

MCMC diagnostics What’s the difference between an MCMC sample and a random i.i.d. sample? Burn-in, autocorrelation, mixing Effective sample size “these diagnostics only find obvious, gross, embarrassing problems that jump out of simple plots.” Increasing "mcmc.interval" will decrease autocorrelation and solve most of your problems. I usually use 5x the number of edges as the interval.

MCMC diagnostics Good: Bad:

Markov chain maximum likelihood estimation Want to maximize the likelihood function with respect to theta Equivalent to maximizing the log-likelihood function: Equivalent to maximizing the log-likelihood minus a constant. We’re free to choose l(θ0) as the constant

Approximating the log-likelihood Augment it with an arbitrary θ0 Approximate this using an MCMC sample We can now maximize this w.r.t. θ

Approximating the normalizing constant Algebra happens... Approximate the expectation with a sample average, where the sample uses the known parameter θ0

Approximating the normalizing constant The algebra: pull the summation out multiply top and bottom by the same thing to get prob given theta0 write sum as expectation

MCMLE algorithm Initialize θ0 with some value Repeat until converge: Get MCMC sample of m networks with θ0 Maximize this function w.r.t. θ Update θ0 to the maximized θ

MCMLE algorithm How do we initialize θ0 ? How big of an MCMC sample do we need? How can we tell when it converges? Since we are using an approximate likelihood, how does that affect our standard error?