Lecture 3 Topic 2: Distributions, hypothesis testing, and sample size determination

Similar documents
Topic 2: Distributions, hypothesis testing, and sample size determination

Practice Final Exam (corrected formulas, 12/10 11AM)

Efficient Estimators for Population Variance using Auxiliary Information

ESTIMATION AND TESTING

t = s D Overview of Tests Two-Sample t-test: Independent Samples Independent Samples t-test Difference between Means in a Two-sample Experiment

Review - Week 10. There are two types of errors one can make when performing significance tests:

Some Improved Estimators for Population Variance Using Two Auxiliary Variables in Double Sampling

The Poisson Process Properties of the Poisson Process

CS344: Introduction to Artificial Intelligence

Chapter 8. Simple Linear Regression

Solution set Stat 471/Spring 06. Homework 2

Least Squares Fitting (LSQF) with a complicated function Theexampleswehavelookedatsofarhavebeenlinearintheparameters

Competitive Facility Location Problem with Demands Depending on the Facilities

-distributed random variables consisting of n samples each. Determine the asymptotic confidence intervals for

The Signal, Variable System, and Transformation: A Personal Perspective

Reliability Analysis. Basic Reliability Measures

Simple Linear Regression Analysis

Fault Tolerant Computing. Fault Tolerant Computing CS 530 Probabilistic methods: overview

Midterm Exam. Tuesday, September hour, 15 minutes

Reaction Time VS. Drug Percentage Subject Amount of Drug Times % Reaction Time in Seconds 1 Mary John Carl Sara William 5 4

(1) Cov(, ) E[( E( ))( E( ))]

Continuous Time Markov Chains

14. Poisson Processes

Speech, NLP and the Web

θ = θ Π Π Parametric counting process models θ θ θ Log-likelihood: Consider counting processes: Score functions:

Analysis of a Stochastic Lotka-Volterra Competitive System with Distributed Delays

Solution. The straightforward approach is surprisingly difficult because one has to be careful about the limits.

Introduction to Hypothesis Testing

Determination of Antoine Equation Parameters. December 4, 2012 PreFEED Corporation Yoshio Kumagae. Introduction

The ray paths and travel times for multiple layers can be computed using ray-tracing, as demonstrated in Lab 3.

Real-Time Systems. Example: scheduling using EDF. Feasibility analysis for EDF. Example: scheduling using EDF

FALL HOMEWORK NO. 6 - SOLUTION Problem 1.: Use the Storage-Indication Method to route the Input hydrograph tabulated below.

Fault Tolerant Computing. Fault Tolerant Computing CS 530 Reliability Analysis

8. Queueing systems lect08.ppt S Introduction to Teletraffic Theory - Fall

Parameters Estimation in a General Failure Rate Semi-Markov Reliability Model

International Journal Of Engineering And Computer Science ISSN: Volume 5 Issue 12 Dec. 2016, Page No.

AML710 CAD LECTURE 12 CUBIC SPLINE CURVES. Cubic Splines Matrix formulation Normalised cubic splines Alternate end conditions Parabolic blending

4. THE DENSITY MATRIX

QR factorization. Let P 1, P 2, P n-1, be matrices such that Pn 1Pn 2... PPA

Final Exam Applied Econometrics

New approach for numerical solution of Fredholm integral equations system of the second kind by using an expansion method

Fundamentals of Speech Recognition Suggested Project The Hidden Markov Model

Chapter Chapter 10 Two-Sample Tests X 1 X 2. Difference Between Two Means: Different data sources Unrelated. Learning Objectives

Reliability Equivalence of a Parallel System with Non-Identical Components

Calibration Approach Based Estimators of Finite Population Mean in Two - Stage Stratified Random Sampling

REVIEW OF SIMPLE LINEAR REGRESSION SIMPLE LINEAR REGRESSION

Lecture 7. Confidence Intervals and Hypothesis Tests in the Simple CLR Model

Interval Estimation. Consider a random variable X with a mean of X. Let X be distributed as X X

BEST PATTERN OF MULTIPLE LINEAR REGRESSION

Chapter 8: Statistical Analysis of Simulated Data

Quantum Mechanics II Lecture 11 Time-dependent perturbation theory. Time-dependent perturbation theory (degenerate or non-degenerate starting state)

Department of Economics University of Toronto

The Linear Regression Of Weighted Segments

Deterioration-based Maintenance Management Algorithm

Cyclone. Anti-cyclone

Least squares and motion. Nuno Vasconcelos ECE Department, UCSD

Pricing Asian Options with Fourier Convolution

The Histogram. Non-parametric Density Estimation. Non-parametric Approaches

Calibration of factor models with equity data: parade of correlations

8 The independence problem

Statistical Inference Procedures

Improved Exponential Estimator for Population Variance Using Two Auxiliary Variables

Econometric Methods. Review of Estimation

Mathematical Formulation

Statistical Estimation

Optimal Eye Movement Strategies in Visual Search (Supplement)

An Efficient Dual to Ratio and Product Estimator of Population Variance in Sample Surveys

Partial Molar Properties of solutions

The textbook expresses the stock price as the present discounted value of the dividend paid and the price of the stock next period.

The textbook expresses the stock price as the present discounted value of the dividend paid and the price of the stock next period.

SYRIAN SEISMIC CODE :

RATIO ESTIMATORS USING CHARACTERISTICS OF POISSON DISTRIBUTION WITH APPLICATION TO EARTHQUAKE DATA

Chapter 3: Maximum-Likelihood & Bayesian Parameter Estimation (part 1)

x z Increasing the size of the sample increases the power (reduces the probability of a Type II error) when the significance level remains fixed.

Lecture Notes Types of economic variables

Nonsynchronous covariation process and limit theorems

Linear Regression. Can height information be used to predict weight of an individual? How long should you wait till next eruption?

Class 13,14 June 17, 19, 2015

The conditional density p(x s ) Bayes rule explained. Bayes rule for a classification problem INF

For the plane motion of a rigid body, an additional equation is needed to specify the state of rotation of the body.

Chapter 11 The Analysis of Variance

Simulation Output Analysis

UNIVERSITY OF OSLO DEPARTMENT OF ECONOMICS

Key words: Fractional difference equation, oscillatory solutions,

Quiz 1- Linear Regression Analysis (Based on Lectures 1-14)

Research Article Centralized Fuzzy Data Association Algorithm of Three-sensor Multi-target Tracking System

Lecture 3. Sampling, sampling distributions, and parameter estimation

European Journal of Mathematics and Computer Science Vol. 5 No. 2, 2018 ISSN

Comments on Discussion Sheet 18 and Worksheet 18 ( ) An Introduction to Hypothesis Testing

F-Tests and Analysis of Variance (ANOVA) in the Simple Linear Regression Model. 1. Introduction

Summary of the lecture in Biostatistics

EE 6885 Statistical Pattern Recognition

Assessing Normality. Assessing Normality. Assessing Normality. Assessing Normality. Normal Probability Plot for Normal Distribution.

European Journal of Mathematics and Computer Science Vol. 5 No. 2, 2018 ISSN

( ) ( ) ( ) ( ) ˆ ˆ ˆ 1. ± n. x ± Where. s ± n Z E. n = x x. = n. STAT 362 Statistics For Management II Formulas. Sample Mean. Sampling Proportions

FORCED VIBRATION of MDOF SYSTEMS

1 Solution to Problem 6.40

Analyzing Two-Dimensional Data. Analyzing Two-Dimensional Data

Use of Non-Conventional Measures of Dispersion for Improved Estimation of Population Mean

Discrete Mathematics and Probability Theory Fall 2016 Seshia and Walrand DIS 10b

Transcription:

Lecure 3 Topc : Drbuo, hypohe eg, ad ample ze deermao The Sude - drbuo Coder a repeaed drawg of ample of ze from a ormal drbuo of mea. For each ample, compue,,, ad aoher ac,, where: The ac he devao of a ormal varable from hypohezed mea meaured adard error u. For ay gve value of, cr alway larger ha Z cr. Th he prce we pay for beg ucera abou he populao varace

Cofdece lm baed o ample ac Takg o accou he mperfec formao provded by amplg, he emaed value of ay populao parameer λ ake he geeral form: λ Emaed λ ± Crcal Value * Sadard error of he emaed λ So, for a populao mea emaed va a ample mea: ±, The ac drbued abou accordg o he drbuo, afyg: P -, +, - The wo erm o eher de repree he lower ad upper - cofdece lm of he mea. The erval bewee hee erm called he cofdece erval CI. - CI for [ -, +, ], Example: Daa e of 4 barley mal exrac value 75.94,.7 / 4 0.379. By -able or R: 0.05,3. 6, 95% CI for 75.94 ±.60.379 75.94 ± 0.7 [75.3, 76.65] If we repeaedly drew radom ample of ze 4 from he populao ad coruced a 95% CI for each, we would expec 95% of hoe erval 9 ou of 0 o coa he rue mea. True mea

3

Hypohe eg ad power Example: Daa e of 4 barley mal exrac value 75.94,.7 / 4 0.379.. Chooe a ull hypohe: Te H 0 : 78 veru H : 78.. Chooe a gfcace level: Ag 0.05. 3. Calculae he e ac: 75.94 78.00 6.8 0.379 4. Compare he abolue value of he e ac o he crcal ac: - 6.8 >.6 5. Sce he abolue value of he e ac larger, we rejec H 0. Th equvale o calculag a 95% cofdece erval aroud. Sce 78 H 0 o wh he 95% CI [75.3, 76.65], we rejec H 0. H 0 rejeced o rejeced rue Type I error Correc deco fale Correc deco Type II error gfcace level Type I error rae he probably of correcly rejecg a rue H 0 β Type II error rae he probably of falg o rejec a fale H 0 Power β he probably of correcly rejecg a fale H 0 4

Power of a e for a gle ample Power β PZ > Z 0 OR P >, 0 Example: Ug he ame barley daa, wha he power of a e for H 0 : 74.88? Aga, 0.05, r 4, 0.05,3.60, ad 0.3795. Power β P > 75.94 74.88.60 P > -.07 0.85 0.3795 The magude of β deped upo:. The Type I error rae. The acual dace bewee he wo mea uder coderao 3. The umber of obervao à For a gve ad deeco dace, f ay wo of he quae, β, or are pecfed, he hrd deermed. Ue uffce replcao o keep Type I ad Type II error uder her dered lm. 5

Fal o rejec H 0 Rejec H 0 / 74.7 74.88 -.60*0.38 74.88 75.588 74.88 +.60*0.38 H 0 True H 0 Fale β Power 75.94 Fal o accep H Accep H Fg. 3. Type I ad Type II error he Barley daa e. H 0 almo alway rejeced f he ample ze oo large ad almo alway o rejeced f he ample ze oo mall. 6

7

8 Power of he e for he dfferece bewee he mea of wo ample H 0 : - 0, veru: H : - 0 wo-aled e H : - < 0 or H : - > 0 oe-aled e The geeral power formula for boh equal ad uequal ample ze : pooled pooled P P Power > > where pooled a weghed varace: + + pooled ad pooled + I he pecal cae of equal ample ze where, he formula mplfy: pooled + pooled + + + + P P Power pooled > > The varace of he dfferece bewee wo radom varable he um of her varace.e. error alway compoud. The degree of freedom for he crcal / ac are: Geeral cae: - + - For equal ample ze: *-

Sample ze for emag, whe kow ug he Z ac If he populao varace kow, or f dered o emae he cofdece erval erm of he rue populao varace, he Z ac may be ued. Z, ad CI ± Z / y Le d repree he half-legh of he cofdece erval: d Z Z Th ca be rearraged o gve a expreo for : Z d For adard 0.05, Z Z 0.05.96 3.84 d d d d If d, 4 If d 0.5, 6 If d 0.5, 64 Th equao ca be re-expreed erm of he coeffce of varao: Z d Z CV d Example: The CV of yeld ral a our expermeal ao are ever greaer ha 5%. How may replcao are eeded o coruc a 95% CI for he rue mea wh a oal legh of o more ha 0% of he rue mea? d 0.0, o d 0.05.96 0.5 / 0.05 34.6 35 9

Sample ze for emag, whe ukow Coder a - % cofdece erval abou ome mea : + -, The half-legh d of h cofdece erval herefore:, d,, Z, d d Se' Two-Sage procedure volve ug a plo udy o emae. Example: A breeder wa o emae he mea hegh of cera maure pla. From a plo udy of 5 pla, he fd ha 0 cm. Wha he requred ample ze, f he wa o have he oal legh of a 95% cofdece erval abou he mea be o loger ha 5 cm? Ug, he ample ze emaed eravely:, d Ial 0.05, Calculaed 5.776.776 0 /.5 3.3 4.96.96 0 /.5 6.5 6.00 64 64.00 64 Thu, wh 64 obervao, oe could emae he rue mea wh a preco of 5 cm, a he gve. Noe ha f we ared wh a Z approxmao, he: Z / d.96 0 /.5 6 0

Sample ze emao for he comparo of wo mea Whe eg he hypohe H 0 :, we ca ake o accou he poble of Type I ad Type II error mulaeouly. To calculae, we eed o kow eher he alerave mea or a lea he mmum dfferece we wh o deec bewee he mea δ -. The approprae formula for compug, he requred umber of obervao from each reame, : / δ Z / + Z β For 0.05 ad β 0.0: Z /.96, Z β 0.846, ad Z / + Z β 7.849 8 If δ, 4 If δ, 6 If δ 0.5, 64 We rarely kow ad mu emae va ample varace: pooled + β, + δ, +, where + pooled Here, emaed eravely. If o emae of avalable, he equao may be expreed erm of he CV ad he dfferece δ a a proporo of he mea: We ca alo defe δ erm of. [/ / δ/] Z / + Z β CV / δ% Z / + Z β Example: Two varee are compared for yeld, wh a prevouly emaed ample varace of.5. How may replcao are eeded o deec a dfferece of.5 o/acre bewee varee? Aume 5% ad β 0%. Approxmae /δ Z / + Z β.5/.5.96+0.846 5.7 Ial df - 0.05, 0.0, Calculaed 6 30.04 0.854 6.8 7 3.037 0.853 6.7 The awer ha here hould be 7 replcao of each varey.

For he ereed: Sample ze o emae populao adard devao The ch-quared drbuo ued o eablh cofdece erval aroud he ample varace a a way of emag he rue, ukow populao varace. The Ch- quare drbuo [ST&D p. 55]. q re F 0.5 0.4 0.3 0. df 4 df 6 df 0. 0.0 3 4 Ch-quare 5 6 The drbuo wh df defed a he um of quare of depede, ormally drbued varable wh zero mea ad u varace. df Z, df Z.96 3., df, df e.g. 3. 84, Z 84, ad.96 3. 84 0.05,

3 Reumg Z If we emae he paramerc mea wh a ample mea, we oba: Z due o: à Th expreo, whch ha a - drbuo, provde a relaohp bewee he ample varace ad he paramerc varace. Cofdece erval for We ca make he followg probablc aeme abou he rao - / :,, P Smple algebrac mapulao of he quae wh he bracke yeld,, P OR,, P

Example: Wha ample ze requred f you wa o oba a emae of ha you are 90% cofde devae o more ha 0% from he rue value of? Tralag h queo o aeme of probably: P 0.8 < / <. 0.90 OR P 0.64 < / <.44 0.90 hu -/, - / - 0.64 AND /, - / -.44 df - / 95% / 5% - - - /- - - /- 0 0.90 0.545 3.4.57 4 40 6.50 0.66 55.8.40 3 30 8.50 0.66 43.8.46 36 35.46 0.64 49.8.4 35 34.66 0.637 48.6.43 Thu a rough emae of he requred ample ze approxmaely 35. 4