Superiority by a Margin Tests for Two Means using Differences

Similar documents
Non-Inferiority Tests for Two Means using Differences

Non-Inferiority Tests for One Mean

Independent Samples T- test

T-test for dependent Samples. Difference Scores. The t Test for Dependent Samples. The t Test for Dependent Samples. s D

Unit 11 Using Linear Regression to Describe Relationships

1) Assume that the sample is an SRS. The problem state that the subjects were randomly selected.

Review of Multiple Regression Richard Williams, University of Notre Dame, Last revised January 13, 2015

A technical guide to 2014 key stage 2 to key stage 4 value added measures

Redesigning Ratings: Assessing the Discriminatory Power of Credit Scores under Censoring

Assessing the Discriminatory Power of Credit Scores

DISTRIBUTED DATA PARALLEL TECHNIQUES FOR CONTENT-MATCHING INTRUSION DETECTION SYSTEMS

REDUCTION OF TOTAL SUPPLY CHAIN CYCLE TIME IN INTERNAL BUSINESS PROCESS OF REAMER USING DOE AND TAGUCHI METHODOLOGY. Abstract. 1.

G*Power 3: A flexible statistical power analysis program for the social, behavioral, and biomedical sciences

DISTRIBUTED DATA PARALLEL TECHNIQUES FOR CONTENT-MATCHING INTRUSION DETECTION SYSTEMS. G. Chapman J. Cleese E. Idle

1 Introduction. Reza Shokri* Privacy Games: Optimal User-Centric Data Obfuscation

Confidence Intervals for Linear Regression Slope

Risk Management for a Global Supply Chain Planning under Uncertainty: Models and Algorithms

Two Dimensional FEM Simulation of Ultrasonic Wave Propagation in Isotropic Solid Media using COMSOL

Queueing systems with scheduled arrivals, i.e., appointment systems, are typical for frontal service systems,

TRADING rules are widely used in financial market as

Control of Wireless Networks with Flow Level Dynamics under Constant Time Scheduling

6. Friction, Experiment and Theory

Ohm s Law. Ohmic relationship V=IR. Electric Power. Non Ohmic devises. Schematic representation. Electric Power

Two-Sample T-Tests Allowing Unequal Variance (Enter Difference)

MBA 570x Homework 1 Due 9/24/2014 Solution

Bidding for Representative Allocations for Display Advertising

Confidence Intervals for the Difference Between Two Means

THE IMPACT OF MULTIFACTORIAL GENETIC DISORDERS ON CRITICAL ILLNESS INSURANCE: A SIMULATION STUDY BASED ON UK BIOBANK ABSTRACT KEYWORDS

Two-Sample T-Tests Assuming Equal Variance (Enter Means)

Exposure Metering Relating Subject Lighting to Film Exposure

Senior Thesis. Horse Play. Optimal Wagers and the Kelly Criterion. Author: Courtney Kempton. Supervisor: Professor Jim Morrow

CHARACTERISTICS OF WAITING LINE MODELS THE INDICATORS OF THE CUSTOMER FLOW MANAGEMENT SYSTEMS EFFICIENCY

INFORMATION Technology (IT) infrastructure management

Acceleration-Displacement Crash Pulse Optimisation A New Methodology to Optimise Vehicle Response for Multiple Impact Speeds

Chapter 10 Stocks and Their Valuation ANSWERS TO END-OF-CHAPTER QUESTIONS

Multi-Objective Optimization for Sponsored Search

Introduction to the article Degrees of Freedom.

QUANTIFYING THE BULLWHIP EFFECT IN THE SUPPLY CHAIN OF SMALL-SIZED COMPANIES

Scheduling of Jobs and Maintenance Activities on Parallel Machines

Health Insurance and Social Welfare. Run Liang. China Center for Economic Research, Peking University, Beijing , China,

Support Vector Machine Based Electricity Price Forecasting For Electricity Markets utilising Projected Assessment of System Adequacy Data.

The Arms Race on American Roads: The Effect of SUV s and Pickup Trucks on Traffic Safety

CASE STUDY ALLOCATE SOFTWARE

Teaching Rank-Based Tests by Emphasizing Structural Similarities to Corresponding Parametric Tests

A note on profit maximization and monotonicity for inbound call centers

TIME SERIES ANALYSIS AND TRENDS BY USING SPSS PROGRAMME

Optical Illusion. Sara Bolouki, Roger Grosse, Honglak Lee, Andrew Ng

POSSIBILITIES OF INDIVIDUAL CLAIM RESERVE RISK MODELING

Profitability of Loyalty Programs in the Presence of Uncertainty in Customers Valuations


Risk-Sharing within Families: Evidence from the Health and Retirement Study

Progress 8 measure in 2016, 2017, and Guide for maintained secondary schools, academies and free schools

A Spam Message Filtering Method: focus on run time

v = x t = x 2 x 1 t 2 t 1 The average speed of the particle is absolute value of the average velocity and is given Distance travelled t

CHAPTER 5 BROADBAND CLASS-E AMPLIFIER

Performance of a Browser-Based JavaScript Bandwidth Test

Morningstar Fixed Income Style Box TM Methodology

The Cash Flow Statement: Problems with the Current Rules

Measuring the Ability of Score Distributions to Model Relevance

Module 8. Three-phase Induction Motor. Version 2 EE IIT, Kharagpur

Tap Into Smartphone Demand: Mobile-izing Enterprise Websites by Using Flexible, Open Source Platforms

Partial optimal labeling search for a NP-hard subclass of (max,+) problems

Queueing Models for Multiclass Call Centers with Real-Time Anticipated Delays

Heat transfer to or from a fluid flowing through a tube

In this paper, we investigate toll setting as a policy tool to regulate the use of roads for dangerous goods

MSc Financial Economics: International Finance. Bubbles in the Foreign Exchange Market. Anne Sibert. Revised Spring Contents

TI-83, TI-83 Plus or TI-84 for Non-Business Statistics

Method of Moments Estimation in Linear Regression with Errors in both Variables J.W. Gillard and T.C. Iles

Pearson's Correlation Tests

A Note on Profit Maximization and Monotonicity for Inbound Call Centers

Unusual Option Market Activity and the Terrorist Attacks of September 11, 2001*

Mixed Method of Model Reduction for Uncertain Systems

Group Mutual Exclusion Based on Priorities

Unobserved Heterogeneity and Risk in Wage Variance: Does Schooling Provide Earnings Insurance?

A Review On Software Testing In SDlC And Testing Tools

FEDERATION OF ARAB SCIENTIFIC RESEARCH COUNCILS

Bio-Plex Analysis Software

Bi-Objective Optimization for the Clinical Trial Supply Chain Management

Problem 1: The Pearson Correlation Coefficient (r) between two variables X and Y can be expressed in several equivalent forms; one of which is

Online story scheduling in web advertising

Towards Control-Relevant Forecasting in Supply Chain Management

Research Article An (s, S) Production Inventory Controlled Self-Service Queuing System

Return on Investment and Effort Expenditure in the Software Development Environment

DUE to the small size and low cost of a sensor node, a

A Life Contingency Approach for Physical Assets: Create Volatility to Create Value

A Resolution Approach to a Hierarchical Multiobjective Routing Model for MPLS Networks

A Comparison of Three Probabilistic Models of Binary Discrete Choice Under Risk

Proceedings of Power Tech 2007, July 1-5, Lausanne

Analysis of Mesostructure Unit Cells Comprised of Octet-truss Structures

Brokerage Commissions and Institutional Trading Patterns

Non-Inferiority Tests for Two Proportions

Growth and Sustainability of Managed Security Services Networks: An Economic Perspective

Is Mark-to-Market Accounting Destabilizing? Analysis and Implications for Policy

Availability of WDM Multi Ring Networks

Achieving Quality Through Problem Solving and Process Improvement

Apigee Edge: Apigee Cloud vs. Private Cloud. Evaluating deployment models for API management

Growth and Sustainability of Managed Security Services Networks: An Economic Perspective

Turbulent Mixing and Chemical Reaction in Stirred Tanks

RISK MANAGEMENT POLICY

January 21, Abstract

Transcription:

Chapter 448 Superiority by a Margin Tet for Two Mean uing Difference Introduction Thi procedure compute power and ample ize for uperiority by a margin tet in two-ample deign in which the outcome i a continuou normal random variable. Meaurement are made on individual that have been randomly aigned to one of two group. Thi i ometime referred to a a parallel-group deign. Thi deign i ued in ituation uch a the comparion of the income level of two region, the nitrogen content of two lake, or the effectivene of two drug. The two-ample t-tet i commonly ued with thi ituation. When the variance of the two group are unequal, Welch t-tet may be ued. When the data are not normally ditributed, the Mann-Whitney (Wilcoxon igned-rank) U tet may be ued. The detail of ample ize calculation for the two-ample deign are preented in the Two-Sample T-Tet chapter and they will not be duplicated here. Thi chapter only dicue thoe change neceary for non-inferiority and uperiority (or non-zero null) tet. Sample ize formula for non-inferiority and uperiority tet of two mean are preented in Chow et al. (003) page 57-59. The Statitical Hypothee Both non-inferiority and uperiority tet are example of directional (one-ided) tet and their power and ample ize could be calculated uing the Two-Sample T-Tet procedure. However, at the urging of our uer, we have developed thi module, which provide the input and output in format that are convenient for thee type of tet. Thi ection will review the pecific of non-inferiority and uperiority teting. Remember that in the uual t-tet etting, the null (H0) and alternative (H) hypothee for one-ided tet are defined a H 0 :µ µ D veru H :µ µ Rejecting thi tet implie that the mean difference i larger than the value D. Thi tet i called an upper-tailed tet becaue it i rejected in ample in which the difference between the ample mean i larger than D. Following i an example of a lower-tailed tet. H 0 :µ µ D veru H :µ µ on-inferiority and uperiority tet are pecial cae of the above directional tet. It will be convenient to adopt the following pecialized notation for the dicuion of thee tet. > D < D 448-

Parameter PASS Input/Output Interpretation µ ot ued Mean of population. Population i aumed to conit of thoe who have received the new treatment. µ ot ued Mean of population. Population i aumed to conit of thoe who have received the reference treatment. M SM Margin of uperiority. Thi i a tolerance value that define the magnitude of difference that i required for practical importance. Thi may be thought of a the mallet difference from the reference that i conidered to be different. δ D True difference. Thi i the value of µ µ, the difference between the mean. Thi i the value at which the power i calculated. ote that the actual value of µ and µ are not needed. Only their difference i needed for power and ample ize calculation. Superiority Tet A uperiority by a margin tet tet that the treatment mean i better than the reference mean by more than the uperiority margin. The actual direction of the hypothei depend on the repone variable being tudied. Cae : High Value Good In thi cae, higher value are better. The hypothee are arranged o that rejecting the null hypothei implie that the treatment mean i greater than the reference mean by at leat the margin of uperiority. The value of δ mut be greater than M. The following are equivalent et of hypothee. S H 0 : µ µ + M S veru H : µ µ + M S > H 0 : µ µ M S veru H : µ µ > M S H 0 :δ M S veru H :δ > M S Cae : High Value Bad In thi cae, lower value are better. The hypothee are arranged o that rejecting the null hypothei implie that the treatment mean i le than the reference mean by at leat the margin of uperiority. The value of δ mut be le than ε. The following are equivalent et of hypothee. H : µ µ veru H : µ < µ M S 0 M S H 0 : µ µ M S veru H : µ µ < M S H :δ M S veru H :δ < M S 0 448-

Example A uperiority tet example will et the tage for the dicuion of the terminology that follow. Suppoe that a tet i to be conducted to determine if a new cancer treatment ubtantially improve mean bone denity. The adjuted mean bone denity (AMBD) in the population of interet i 0.00300 gm/cm with a tandard deviation of 0.000300 gm/cm. Clinician decide that if the treatment increae AMBD by more than 5% (0.0005 gm/cm), it provide a ignificant health benefit. The hypothei of interet i whether the mean AMBD in the treated group i more than 0.0005 above that of the reference group. The tatitical tet will be et up o that if the null hypothei i rejected, the concluion will be that the new treatment i uperior. The value 0.0005 gm/cm i called the margin of uperiority. Tet Statitic Thi ection decribe the tet tatitic that are available in thi procedure. Two-Sample T-Tet Under the null hypothei, thi tet aume that the two group of data are imple random ample from a ingle population of normally-ditributed value that all have the ame mean and variance. Thi aumption implie that the data are continuou and their ditribution i ymmetric. The calculation of the tet tatitic for the cae when higher repone value are good i a follow. where t df ( X X) X X ε X k k X i k ki X X ( X i X) + ( X i X ) i i + + df + The null hypothei i rejected if the computed p-value i le than a pecified level (uually 0.05). Otherwie, no concluion can be reached. Welch T-Tet Welch (938) propoed the following tet when the two variance are not aumed to be equal. t * f ( X X ) * X X ε 448-3

where * X X ( X i X) ( ) i + i ( X i X ) ( ) f + 4 + ( ) ( ) 4 ( X i X) i, ( X i X ) i Mann-Whitney U Tet Thi tet i the nonparametric ubtitute for the equal-variance t-tet. Two key aumption are that the ditribution are at leat ordinal and that they are identical under H0. Thi mean that tie (repeated value) are not acceptable. When tie are preent, you can ue approximation, but the theoretic reult no longer hold. The Mann-Whitney tet tatitic i defined a follow in Gibbon (985). where W ( ) Rank X k k W ( + + ) + C z The rank are determined after combining the two ample. The tandard deviation i calculated a W W 3 ( ti ti ) ( + + ) i ( + )( + ) where t i i the number of obervation tied at value one, t i the number of obervation tied at ome value two, and o forth. The correction factor, C, i 0.5 if the ret of the numerator i negative or -0.5 otherwie. The value of z i then compared to the normal ditribution. 448-4

Computing the Power Standard Deviation Equal When σ σ σ, the power of the t tet i calculated a follow.. Find t α uch that Tdf ( tα ) α, where Tdf ( tα ) df +.. Calculate: σ σ x + ε δ 3. Calculate the noncentrality parameter: λ σ 4. Calculate: Power Tdf,λ ( tα ), where T ( x) x i the area under a central-t curve to the left of x and df,λ i the area to the left of x under a noncentral-t curve with degree of freedom df and noncentrality parameter λ. Standard Deviation Unequal Thi cae often recommend Welch tet. When σ σ, the power i calculated a follow. σ σ. Calculate: σ +.. Calculate: f x 4 σ x - σ 4 ( +) + σ 4 ( +) which i the adjuted degree of freedom. Often, thi i rounded to the next highet integer. ote that thi i not the value of f ued in the computation of the actual tet. Intead, thi i the expected value of f. 3. Find t α uch that Tf ( tα ) α, where Tf ( tα ) degree of freedom. ε 4. Calculate: λ, the noncentrality parameter. σ x 5. Calculate: Power Tf,λ ( tα ) ( ), where T x degree of freedom f and noncentrality parameter λ. i the area to the left of x under a central-t curve with f f,λ i the area to the left of x under a noncentral-t curve with onparametric Adjutment When uing the Mann-Whitney tet rather than the t tet, reult by Al-Sunduqchi and Guenther (990) indicate that power calculation for the Mann-Whitney tet may be made uing the tandard t tet formulation with a imple adjutment to the ample ize. The ize of the adjutment depend on the actual ditribution of the data. They give ample ize adjutment factor for four ditribution. Thee are for uniform, /3 for double exponential, 9 / π for logitic, and π / 3 for normal ditribution. 448-5

Procedure Option Thi ection decribe the option that are pecific to thi procedure. Thee are located on the Deign tab. For more information about the option of other tab, go to the Procedure Window chapter. Deign Tab The Deign tab contain mot of the parameter and option that you will be concerned with. Solve For Solve For Thi option pecifie the parameter to be calculated from the value of the other parameter. Under mot condition, you would elect either Power or Sample Size (). Select Sample Size () when you want to determine the ample ize needed to achieve a given power and alpha. Select Power when you want to calculate the power of an experiment that ha already been run. Tet Higher Mean Are Thi option define whether higher value of the repone variable are to be conidered better or wore. The choice here determine the direction of the tet. If Higher Mean Are Better the null hypothei i Diff SM and the alternative hypothei i Diff > SM. If Higher Mean Are Wore the null hypothei i Diff -SM and the alternative hypothei i Diff < -SM. onparametric Adjutment (Mann-Whitney Tet) Thi option make appropriate ample ize adjutment for the Mann-Whitney tet. Reult by Al-Sunduqchi and Guenther (990) indicate that power calculation for the Mann-Whitney tet may be made uing the tandard t tet formulation with a imple adjutment to the ample ize. The ize of the adjutment depend upon the actual ditribution of the data. They give ample ize adjutment factor for four ditribution. Thee are for the uniform ditribution, /3 for the double exponential ditribution, 9 / π for the logitic ditribution, and π / 3 for the normal ditribution. The option are a follow: Ignore Do not make a Mann-Whitney adjutment. Thi indicate that you want to analyze a t tet, not the Wilcoxon tet. Uniform Make the Mann-Whitney ample ize adjutment auming the uniform ditribution. Since the factor i one, thi option perform the ame function a Ignore. It i included for completene. Double Exponential Make the Mann-Whitney ample ize adjutment auming that the data actually follow the double exponential ditribution. 448-6

Logitic Make the Mann-Whitney ample ize adjutment auming that the data actually follow the logitic ditribution. ormal Make the Mann-Whitney ample ize adjutment auming that the data actually follow the normal ditribution. Power and Alpha Power Thi option pecifie one or more value for power. Power i the probability of rejecting a fale null hypothei, and i equal to one minu Beta. Beta i the probability of a type-ii error, which occur when a fale null hypothei i not rejected. In thi procedure, a type-ii error occur when you fail to reject the null hypothei of inferiority when the null hypothei hould be rejected. Value mut be between zero and one. Hitorically, the value of 0.80 (Beta 0.0) wa ued for power. ow, 0.90 (Beta 0.0) i alo commonly ued. A ingle value may be entered here or a range of value uch a 0.8 to 0.95 by 0.05 may be entered. Alpha Thi option pecifie one or more value for the probability of a type-i error. A type-i error occur when a true null hypothei i rejected. In thi procedure, a type-i error occur when you reject the null hypothei of inferiority when in fact the mean i not non-inferior. Value mut be between zero and one. Hitorically, the value of 0.05 ha been ued for alpha. Thi mean that about one tet in twenty will falely reject the null hypothei. You hould pick a value for alpha that repreent the rik of a type-i error you are willing to take in your experimental ituation. You may enter a range of value uch a 0.0 0.05 0.0 or 0.0 to 0.0 by 0.0. Sample Size (When Solving for Sample Size) Group Allocation Select the option that decribe the contraint on or or both. The option are Equal ( ) Thi election i ued when you wih to have equal ample ize in each group. Since you are olving for both ample ize at once, no additional ample ize parameter need to be entered. Enter, olve for Select thi option when you wih to fix at ome value (or value), and then olve only for. Pleae note that for ome value of, there may not be a value of that i large enough to obtain the deired power. Enter, olve for Select thi option when you wih to fix at ome value (or value), and then olve only for. Pleae note that for ome value of, there may not be a value of that i large enough to obtain the deired power. Enter R /, olve for and For thi choice, you et a value for the ratio of to, and then PASS determine the needed and, with thi ratio, to obtain the deired power. An equivalent repreentation of the ratio, R, i 448-7

R *. Enter percentage in Group, olve for and For thi choice, you et a value for the percentage of the total ample ize that i in Group, and then PASS determine the needed and with thi percentage to obtain the deired power. (Sample Size, Group ) Thi option i diplayed if Group Allocation Enter, olve for i the number of item or individual ampled from the Group population. mut be. You can enter a ingle value or a erie of value. (Sample Size, Group ) Thi option i diplayed if Group Allocation Enter, olve for i the number of item or individual ampled from the Group population. mut be. You can enter a ingle value or a erie of value. R (Group Sample Size Ratio) Thi option i diplayed only if Group Allocation Enter R /, olve for and. R i the ratio of to. That i, R /. Ue thi value to fix the ratio of to while olving for and. Only ample ize combination with thi ratio are conidered. i related to by the formula: where the value [Y] i the next integer Y. [R ], For example, etting R.0 reult in a Group ample ize that i double the ample ize in Group (e.g., 0 and 0, or 50 and 00). R mut be greater than 0. If R <, then will be le than ; if R >, then will be greater than. You can enter a ingle or a erie of value. Percent in Group Thi option i diplayed only if Group Allocation Enter percentage in Group, olve for and. Ue thi value to fix the percentage of the total ample ize allocated to Group while olving for and. Only ample ize combination with thi Group percentage are conidered. Small variation from the pecified percentage may occur due to the dicrete nature of ample ize. The Percent in Group mut be greater than 0 and le than 00. You can enter a ingle or a erie of value. 448-8

Sample Size (When ot Solving for Sample Size) Group Allocation Select the option that decribe how individual in the tudy will be allocated to Group and to Group. The option are Equal ( ) Thi election i ued when you wih to have equal ample ize in each group. A ingle per group ample ize will be entered. Enter and individually Thi choice permit you to enter different value for and. Enter and R, where R * Chooe thi option to pecify a value (or value) for, and obtain a a ratio (multiple) of. Enter total ample ize and percentage in Group Chooe thi option to pecify a value (or value) for the total ample ize (), obtain a a percentage of, and then a -. Sample Size Per Group Thi option i diplayed only if Group Allocation Equal ( ). The Sample Size Per Group i the number of item or individual ampled from each of the Group and Group population. Since the ample ize are the ame in each group, thi value i the value for, and alo the value for. The Sample Size Per Group mut be. You can enter a ingle value or a erie of value. (Sample Size, Group ) Thi option i diplayed if Group Allocation Enter and individually or Enter and R, where R *. i the number of item or individual ampled from the Group population. mut be. You can enter a ingle value or a erie of value. (Sample Size, Group ) Thi option i diplayed only if Group Allocation Enter and individually. i the number of item or individual ampled from the Group population. mut be. You can enter a ingle value or a erie of value. R (Group Sample Size Ratio) Thi option i diplayed only if Group Allocation Enter and R, where R *. R i the ratio of to. That i, R / Ue thi value to obtain a a multiple (or proportion) of. i calculated from uing the formula: where the value [Y] i the next integer Y. [R x ], 448-9

For example, etting R.0 reult in a Group ample ize that i double the ample ize in Group. R mut be greater than 0. If R <, then will be le than ; if R >, then will be greater than. You can enter a ingle value or a erie of value. Total Sample Size () Thi option i diplayed only if Group Allocation Enter total ample ize and percentage in Group. Thi i the total ample ize, or the um of the two group ample ize. Thi value, along with the percentage of the total ample ize in Group, implicitly define and. The total ample ize mut be greater than one, but practically, mut be greater than 3, ince each group ample ize need to be at leat. You can enter a ingle value or a erie of value. Percent in Group Thi option i diplayed only if Group Allocation Enter total ample ize and percentage in Group. Thi value fixe the percentage of the total ample ize allocated to Group. Small variation from the pecified percentage may occur due to the dicrete nature of ample ize. The Percent in Group mut be greater than 0 and le than 00. You can enter a ingle value or a erie of value. Effect Size Mean Difference SM (Superiority Margin) Thi i the magnitude of the margin of uperiority. It mut be entered a a poitive number. When higher mean are better, thi value i the ditance above the reference mean that i required to be conidered uperior. When higher mean are wore, thi value i the ditance below the reference mean that i required to be conidered uperior. D (True Difference, Trt Mean Ref Mean) Thi i the actual difference between the treatment mean and the reference mean at which the power i calculated. When higher mean are better, thi value hould be greater than SM. When higher mean are wore, thi value hould be negative and greater in magnitude than SM. Effect Size Standard Deviation S and S (Standard Deviation) Thee option pecify the value of the tandard deviation for each group. When the S i et to S, the EQUAL VARIACE tet i ued and only S need to be pecified. The value of S will be ued for S. Otherwie, the UEQUAL VARIACE tet i ued (even if the value entered for S equal S). When thee value are not known, you mut upply etimate of them. Pre the SD button to diplay the Standard Deviation Etimator window. Thi procedure will help you find appropriate value for the tandard deviation. 448-0

Example Power Analyi Suppoe that a tet i to be conducted to determine if a new cancer treatment improve bone denity. The adjuted mean bone denity (AMBD) in the population of interet i 0.00300 gm/cm with a tandard deviation of 0.000300 gm/cm. Clinician decide that if the treatment increae AMBD by more than 5% (0.0005 gm/cm), it generate a ignificant health benefit. They alo want to conider what would happen if the margin of uperiority i et to.5% (0.0000575 gm/cm). The analyi will be a non- zero null tet uing the t-tet at the 0.05 ignificance level. Power to be calculated auming that the new treatment ha 7.5% improvement on AMBD. Several ample ize between 0 and 800 will be analyzed. The reearcher want to achieve a power of at leat 90%. All number have been multiplied by 0000 to make the report and plot eaier to read. Setup Thi ection preent the value of each of the parameter needed to run thi example. Firt, from the PASS Home window, load the procedure window by expanding Mean, then Two Independent Mean, then clicking on Superiority by a Margin, and then clicking on. You may then make the appropriate entrie a lited below, or open Example by going to the File menu and chooing Open Example Template. Option Value Deign Tab Solve For... Power Higher Mean Are... Better onparametric Adjutment... Ignore Alpha... 0.05 Group Allocation... Equal ( ) Sample Size Per Group... 0 50 00 00 300 500 600 800 SM (Superiority Margin)... 0.575.5 D (True Difference)....75 S (Standard Deviation Group )... 3 S (Standard Deviation Group )... S Annotated Output Click the Calculate button to perform the calculation and generate the following output. umeric Reult and Plot umeric Reult for Superiority Tet (H0: Diff SM; H: Diff > SM) Higher Mean are Better Tet Statitic: T-Tet 0.553 0 0 0 0.575.75 3.0 3.0 0.05 0.4754 50 50 00 0.575.75 3.0 3.0 0.05 0.76957 00 00 00 0.575.75 3.0 3.0 0.05 0.96885 00 00 400 0.575.75 3.0 3.0 0.05 0.9968 300 300 600 0.575.75 3.0 3.0 0.05 0.99998 500 500 000 0.575.75 3.0 3.0 0.05.00000 600 600 00 0.575.75 3.0 3.0 0.05.00000 800 800 600 0.575.75 3.0 3.0 0.05 0.0603 0 0 0.50.75 3.0 3.0 0.05 0.560 50 50 00.50.75 3.0 3.0 0.05 (report continue) 448-

Report Definition Power i the probability of rejecting a fale null hypothei. and are the number of item ampled from each population. i the total ample ize, +. SM i the magnitude of the margin of uperiority. Since higher mean are better, thi value i poitive and i the ditance above the reference mean that i required to be conidered uperior. D i the mean difference at which the power i computed. D Mean - Mean, or Treatment Mean - Reference Mean. S and S are the aumed population tandard deviation for group and, repectively. Alpha i the probability of rejecting a true null hypothei. Summary Statement Group ample ize of 0 and 0 achieve 3% power to detect uperiority uing a one-ided, two-ample t-tet. The margin of uperiority i 0.575. The true difference between the mean i aumed to be.75. The ignificance level (alpha) of the tet i 0.0500. The data are drawn from population with tandard deviation of 3.000 and 3.000. Chart Section 448-

The above report how that for SM.5, the ample ize neceary to obtain 90% power i about 600 per group. However, if SM 0.575, the required ample ize i only about 80 per group. 448-3

Example Finding the Sample Size Continuing with Example, the reearcher want to know the exact ample ize for each value of SM to achieve 90% power. Setup Thi ection preent the value of each of the parameter needed to run thi example. Firt, from the PASS Home window, load the procedure window by expanding Mean, then Two Independent Mean, then clicking on Superiority by a Margin, and then clicking on. You may then make the appropriate entrie a lited below, or open Example by going to the File menu and chooing Open Example Template. Option Value Deign Tab Solve For... Sample Size Higher Mean Are... Better onparametric Adjutment... Ignore Power... 0.90 Alpha... 0.05 Group Allocation... Equal ( ) SM (Superiority Margin)... 0.575.5 D (True Difference)....75 S (Standard Deviation Group )... 3 S (Standard Deviation Group )... S Output Click the Calculate button to perform the calculation and generate the following output. umeric Reult umeric Reult for Superiority Tet (H0: Diff SM; H: Diff > SM) Higher Mean are Better Tet Statitic: T-Tet Target Actual Power Power SM D S S Alpha 0.90 0.90004 44 44 88 0.6.7 3.0 3.0 0.05 0.90 0.90036 573 573 46..7 3.0 3.0 0.05 Thi report how the exact ample ize requirement for each value of SM. Example 3 Validation Thi procedure ue the ame mechanic a the on-inferiority Tet for Two Mean uing Difference procedure. We refer the uer to Example 3 and 4 of Chapter 450 for the validation. 448-4