Topic 5 - Joint distributions and the CLT

Similar documents
This is a good time to refresh your memory on double-integration. We will be using this skill in the upcoming lectures.

Probability Model for 2 RV s

Exam 2 is Tue Nov 21. Bring a pencil and a calculator. Discuss similarity to exam1. HW3 is due Tue Dec 5.

Will Monroe July 21, with materials by Mehran Sahami and Chris Piech. Joint Distributions

This is a good time to refresh your memory on double-integration. We will be using this skill in the upcoming lectures.

Today s outline: pp

Lecture 8: Jointly distributed random variables

Joint probability distributions

Multivariate probability distributions

Probability and Statistics for Final Year Engineering Students

Page 129 Exercise 5: Suppose that the joint p.d.f. of two random variables X and Y is as follows: { c(x. 0 otherwise. ( 1 = c. = c

Pairs of a random variable

Statistics I 2011/2012 Notes about the third Computer Class: Simulation of samples and goodness of fit; Central Limit Theorem; Confidence intervals.

The Normal Distribution & z-scores

Central Limit Theorem Sample Means

The Normal Distribution & z-scores

CS 112: Computer System Modeling Fundamentals. Prof. Jenn Wortman Vaughan April 21, 2011 Lecture 8

Lecture 6: Chapter 6 Summary

The Normal Distribution & z-scores

ECE353: Probability and Random Processes. Lecture 11- Two Random Variables (II)

Chapter 1. Looking at Data-Distribution

Kernel Density Estimation (KDE)

4.3 The Normal Distribution

Chapter 6 Normal Probability Distributions

14.30 Introduction to Statistical Methods in Economics Spring 2009

An Introduction to PDF Estimation and Clustering

Math 227 EXCEL / MEGASTAT Guide

LAB #2: SAMPLING, SAMPLING DISTRIBUTIONS, AND THE CLT

Today. Lecture 4: Last time. The EM algorithm. We examine clustering in a little more detail; we went over it a somewhat quickly last time

Recursive Estimation

Probability Models.S4 Simulating Random Variables

CS 2316 Individual Homework 5 Joint Probability Out of 100 points

Learner Expectations UNIT 1: GRAPICAL AND NUMERIC REPRESENTATIONS OF DATA. Sept. Fathom Lab: Distributions and Best Methods of Display

Direction Fields; Euler s Method

1 Overview of Statistics; Essential Vocabulary

Comprehensive Practice Handout MATH 1325 entire semester

MIDTERM. Section: Signature:

CHAPTER 8: INTEGRALS 8.1 REVIEW: APPROXIMATING INTEGRALS WITH RIEMANN SUMS IN 2-D

Unit 5: Estimating with Confidence

Chapter 3 - Displaying and Summarizing Quantitative Data

Chapter 6 The Standard Deviation as Ruler and the Normal Model

Behavior of the sample mean. varx i = σ 2

Acquisition Description Exploration Examination Understanding what data is collected. Characterizing properties of data.

Chapters 5-6: Statistical Inference Methods

STA 570 Spring Lecture 5 Tuesday, Feb 1

Generative and discriminative classification techniques

Lecture 09: Continuous RV. Lisa Yan July 16, 2018

Integration. Example Find x 3 dx.

LAB 1 INSTRUCTIONS DESCRIBING AND DISPLAYING DATA

Density Curve (p52) Density curve is a curve that - is always on or above the horizontal axis.

height VUD x = x 1 + x x N N 2 + (x 2 x) 2 + (x N x) 2. N

Chapter 1. Introduction

Learning Objectives. Continuous Random Variables & The Normal Probability Distribution. Continuous Random Variable

MATH : EXAM 3 INFO/LOGISTICS/ADVICE

Quarter 3 Review - Honors

Chapter 5: Joint Probability Distributions and Random

To complete the computer assignments, you ll use the EViews software installed on the lab PCs in WMC 2502 and WMC 2506.

Visualizing univariate data 1

Ch6: The Normal Distribution

Chapter 3 Transformations of Graphs and Data

We use non-bold capital letters for all random variables in these notes, whether they are scalar-, vector-, matrix-, or whatever-valued.

Chapter 2 Exploring Data with Graphs and Numerical Summaries

Math Exam 2a. 1) Take the derivatives of the following. DO NOT SIMPLIFY! 2 c) y = tan(sec2 x) ) b) y= , for x 2.

STAT 135 Lab 1 Solutions

Introduction to RStudio

Distributions of Continuous Data

Chapter 6: DESCRIPTIVE STATISTICS

Chapter 6: Simulation Using Spread-Sheets (Excel)

ISyE 6416: Computational Statistics Spring Lecture 13: Monte Carlo Methods

BIO 360: Vertebrate Physiology Lab 9: Graphing in Excel. Lab 9: Graphing: how, why, when, and what does it mean? Due 3/26

MATH11400 Statistics Homepage

1/12/2009. Image Elements (Pixels) Image Elements (Pixels) Digital Image. Digital Image =...

MULTI-DIMENSIONAL MONTE CARLO INTEGRATION

BIOSTATISTICS LABORATORY PART 1: INTRODUCTION TO DATA ANALYIS WITH STATA: EXPLORING AND SUMMARIZING DATA

Intro to Probability Instructor: Alexandre Bouchard

Lecture 3 Questions that we should be able to answer by the end of this lecture:

Statistics Lecture 6. Looking at data one variable

Biostatistics & SAS programming. Kevin Zhang

Page 1. Graphical and Numerical Statistics

How to Make Graphs in EXCEL

Lecture 3 Questions that we should be able to answer by the end of this lecture:

Section 6.2: Generating Discrete Random Variates

Organizing and Summarizing Data

Triple Integrals. MATH 311, Calculus III. J. Robert Buchanan. Fall Department of Mathematics. J. Robert Buchanan Triple Integrals

Parameter Estimation. Learning From Data: MLE. Parameter Estimation. Likelihood. Maximum Likelihood Parameter Estimation. Likelihood Function 12/1/16

CHAPTER 6. The Normal Probability Distribution

Univariate descriptives

1 RefresheR. Figure 1.1: Soy ice cream flavor preferences

Announcements. Topics: To Do:

Nonparametric Density Estimation

C ontent descriptions

Key: 5 9 represents a team with 59 wins. (c) The Kansas City Royals and Cleveland Indians, who both won 65 games.

CS 237 Fall 2018, Homework 08 Solution

= f (a, b) + (hf x + kf y ) (a,b) +

Regression III: Advanced Methods

Math 120 Introduction to Statistics Mr. Toner s Lecture Notes 3.1 Measures of Central Tendency

Quick Start Guide. Version R94. English

Package simed. November 27, 2017

(c) 0 (d) (a) 27 (b) (e) x 2 3x2

These slides follow closely the (English) course textbook Pattern Recognition and Machine Learning by Christopher Bishop

Transcription:

Topic 5 - Joint distributions and the CLT Joint distributions Calculation of probabilities, mean and variance Expectations of functions based on joint distributions Central Limit Theorem Sampling distributions Of the mean Of totals 1

Often times, we are interested in more than one random variable at a time. For example, what is the probability that a car will have at least one engine problem and at least one blowout during the same week? X = # of engine problems in a week Y = # of blowouts in a week P(X 1, Y 1) is what we are looking for To understand these sorts of probabilities, we need to develop joint distributions. 2

Discrete distributions A discrete joint probability mass function is given by f(x,y) = P(X = x, Y = y) where 1. f( x, y) 0 for all x, y 2. f( x, y) 1 all ( xy, ) 3. P(( X, Y) A) f( x, y) all ( xy, ) A 4. EhXY ( (, )) hx (, y) f( x, y) all ( xy, ) 3

Return to the car example Consider the following joint pmf for X and Y X\Y 0 1 2 3 4 0 1/2 1/16 1/32 1/32 1/32 1 1/16 1/32 1/32 1/32 1/32 2 1/32 1/32 1/32 1/32 1/32 P(X 1, Y 1) = P(X 1) = E(X + Y) = 4

Joint to marginals The probability mass functions for X and Y individually (called marginals) are given by f ( x) f( x, y), f ( y) f( x, y) X all y Y all x Returning to the car example: f X (x) = f Y (y) = E(X) = E(Y) = 5

Continuous distributions A joint probability density function for two continuous random variables, (X,Y), has the following four properties: 1. f( x, y) 0 for all x, y 2. f( x, y) dxdy 1 - - 3. P(( X, Y) A) f( x, y) dxdy A 4. EhXY ( (, )) hx (, y) f( x, ydxdy ) - - 6

Continuous example Consider the following joint pdf: 2 (1 3 ) x y f( x, y) 0 x 2, 0 y 1 4 Show condition 2 (total volume is 1) holds on your own. Show P(0 < X < 1, ¼ < Y < ½) = 23/512 11/2 2 x(1 3 y ) P(0 x1,1/ 4 y1/ 2) dydx 4 01/4 1 1 3 y1/2 y1/4 0 0 1/ 4 xy [ y] dx1/ 4 x[5 / 8 17 / 64] dx 1 23/ 256 xdx 23/ 256[ x / 2] 23/ 256[1/ 2 0] 23/ 512 0 2 x1 x0 7

Joint to marginals The marginal pdfs for X and Y can be found by f ( x) f( x, y) dy, f ( y) f( x, y) dx X For the previous example, find f X (x) and f Y (y). Y 1 2 x(1 3 y ) 3 y1 fx( x) dy = x/ 4[ y y ] y0 = x/ 4[2 0] x/ 2 4 0 2 2 2 2 2 x(13 y ) (13 y ) (13 y ) fy ( y) dx = xdx = [ x / 2] 4 4 4 0 0 2 x2 x0 1 3y 2 2 8

Independence of X and Y The random variables X and Y are independent if f(x,y) = f X (x) f Y (y) for all pairs (x,y). For the discrete clunker car example, are X and Y independent? For the continuous example, are X and Y independent? x y x y x y f( x, y) f ( x) f ( y) ( ) 4 2 2 4 2 2 2 (1 3 ) (1 3 ) (1 3 ) x y 9

Sampling distributions We assume that each data value we collect represents a random selection from a common population distribution. The collection of these independent random variables is called a random sample from the distribution. A statistic is a function of these random variables that is used to estimate some characteristic of the population distribution. The distribution of a statistic is called a sampling distribution. The sampling distribution is a key component to making inferences about the population. 10

Statistics used to infer parameters We take samples and calculate statistics to make inferences about the population parameters. Sample Population Mean x Std. Dev. s Variance 2 s 2 Proportion ˆp p 11

StatCrunch example StatCrunch subscriptions are sold for 6 months ($5) or 12 months ($8). From past data, I can tell you that roughly 80% of subscriptions are $5 and 20% are $8. Let X represent the amount in $ of a purchase. E(X) = Var(X) = 12

StatCrunch example continued Now consider the amounts of a random sample of two purchases, X 1, X 2. A natural statistic of interest is X 1 + X 2, the total amount of the purchases. Outcomes X 1 + X 2 5,5 Probability X 1 + X 2 Probability 5,8 8,5 8,8 13

StatCrunch example continued E(X 1 + X 2 ) = E([X 1 + X 2 ] 2 ) = Var(X 1 + X 2 ) = 14

StatCrunch example continued If I have n purchases in a day, what is my expected earnings? the variance of my earnings? the shape of my earnings distribution for large n? Let s experiment by simulating 10,000 days with 100 purchases per day using StatCrunch. 15

Simulation instructions Data > Simulate data > Binomial Specify Rows to be 10000, Columns to be 1, n to be 100 and p to be.2. This will give you a new column called Binomial1 To compute the total for each day, go to Data > Transform data and enter the expression, 8*Binomial1+5*(100-Binomial1). This will add a new column to the data table. Make a histogram and set the bin width to 1 for best results. For the new sum column, do a histogram and a QQ plot. Both should verify normality! StatCrunch 16

Should result in a dataset like this 17

Central Limit Theorem We have just illustrated one of the most important theorems in statistics. As the sample size, n, becomes large the distribution of the sum of a random sample from a distribution with mean and variance 2 converges to a Normal distribution with mean n and variance n 2. A sample size of at least 30 is typically required to use the CLT (arguable in the general statistics community). The amazing part of this theorem is that it is true regardless of the form of the underlying distribution. 18

Airplane example Suppose the weight of an airline passenger has a mean of 150 lbs. and a standard deviation of 25 lbs. What is the probability the combined weight of 100 passengers will exceed the maximum allowable weight of 15,500 lbs? How many passengers should be allowed on the plane if we want this probability to be at most 0.01? 19

What are the probabilities at n = 99? 99*150 14850 99*25 61850 The mean is 2 The variance is The standard deviation is 61875 248.75 PX ( 15500) TOT 0.004487 20

The distribution of the sample means For constant c, E(cY) = ce(y) and Var(cY) = c 2 Var(Y) 1 1 1 2 n n n n Var( X ) Var( x) Var( x) n 2 2 2 The CLT says that for large samples, X is approximately normal with a mean of and a variance of 2 /n. So, the variance of the sample mean decreases with n. 21

What are the probabilities we get a sample average at some level? If the parent population is assumed with a mean of 150 lbs. and a standard deviation of 25 lbs., what s the probability we get a sample average below 141 with a sample size of 30? Talking about the sampling distribution, the mean is 150 and the standard deviation is 25 30 4.5644 22

Sampling distribution applet In StatCrunch, go to the Applets tab and click on sampling distributions. It will demonstrate how any parent distribution will converge to normal with larger, repeated samples. The closer the parent is to symmetrical, the quicker the sampling distribution will converge. The additional file for Topic 5 has discussion and examples on both sampling distributions and joint probability distributions. There are also additional examples of double integration. 23