On Construction of Odd-fractional Factorial Designs

Similar documents
Robustness Experiments with Two Variance Components

( t) Outline of program: BGC1: Survival and event history analysis Oslo, March-May Recapitulation. The additive regression model

TSS = SST + SSE An orthogonal partition of the total SS

John Geweke a and Gianni Amisano b a Departments of Economics and Statistics, University of Iowa, USA b European Central Bank, Frankfurt, Germany

Approximate Analytic Solution of (2+1) - Dimensional Zakharov-Kuznetsov(Zk) Equations Using Homotopy

A NEW TECHNIQUE FOR SOLVING THE 1-D BURGERS EQUATION

In the complete model, these slopes are ANALYSIS OF VARIANCE FOR THE COMPLETE TWO-WAY MODEL. (! i+1 -! i ) + [(!") i+1,q - [(!

MAXIMIN POWER DESIGNS IN TESTING LACK OF FIT Douglas P. Wiens 1

V.Abramov - FURTHER ANALYSIS OF CONFIDENCE INTERVALS FOR LARGE CLIENT/SERVER COMPUTER NETWORKS

Tight results for Next Fit and Worst Fit with resource augmentation

A L A BA M A L A W R E V IE W

GENERATING CERTAIN QUINTIC IRREDUCIBLE POLYNOMIALS OVER FINITE FIELDS. Youngwoo Ahn and Kitae Kim

Delay Dependent Robust Stability of T-S Fuzzy. Systems with Additive Time Varying Delays

Relative controllability of nonlinear systems with delays in control

CS434a/541a: Pattern Recognition Prof. Olga Veksler. Lecture 4

Outline. Energy-Efficient Target Coverage in Wireless Sensor Networks. Sensor Node. Introduction. Characteristics of WSN

[ ] 2. [ ]3 + (Δx i + Δx i 1 ) / 2. Δx i-1 Δx i Δx i+1. TPG4160 Reservoir Simulation 2018 Lecture note 3. page 1 of 5

Single-loop System Reliability-Based Design & Topology Optimization (SRBDO/SRBTO): A Matrix-based System Reliability (MSR) Method

Stability Analysis of Fuzzy Hopfield Neural Networks with Timevarying

New M-Estimator Objective Function. in Simultaneous Equations Model. (A Comparative Study)

FTCS Solution to the Heat Equation

An introduction to Support Vector Machine

Midterm Exam. Thursday, April hour, 15 minutes

HEAT CONDUCTION PROBLEM IN A TWO-LAYERED HOLLOW CYLINDER BY USING THE GREEN S FUNCTION METHOD

Review of Numerical Schemes for Two Point Second Order Non-Linear Boundary Value Problems

Robust and Accurate Cancer Classification with Gene Expression Profiling

DEEP UNFOLDING FOR MULTICHANNEL SOURCE SEPARATION SUPPLEMENTARY MATERIAL

On One Analytic Method of. Constructing Program Controls

SOME NOISELESS CODING THEOREMS OF INACCURACY MEASURE OF ORDER α AND TYPE β

An adaptive approach to small object segmentation

MAXIMIN POWER DESIGNS IN TESTING LACK OF FIT Douglas P. Wiens 1. July 30, 2018

P R = P 0. The system is shown on the next figure:

Track Properities of Normal Chain

Fall 2010 Graduate Course on Dynamic Learning

Dual Approximate Dynamic Programming for Large Scale Hydro Valleys

Chapter 6: AC Circuits

Comparison between the Discrete and Continuous Time Models

CHAPTER 10: LINEAR DISCRIMINATION

Comparison of Differences between Power Means 1

Lecture 11 SVM cont

Normal Random Variable and its discriminant functions

Bernoulli process with 282 ky periodicity is detected in the R-N reversals of the earth s magnetic field

Supporting Information: The integrated Global Temperature change Potential (igtp) and relationships between emission metrics

T q (heat generation) Figure 0-1: Slab heated with constant source 2 = q k

This document is downloaded from DR-NTU, Nanyang Technological University Library, Singapore.

û s L u t 0 s a ; i.e., û s 0

M. Y. Adamu Mathematical Sciences Programme, AbubakarTafawaBalewa University, Bauchi, Nigeria

ABSTRACT KEYWORDS. Bonus-malus systems, frequency component, severity component. 1. INTRODUCTION

Density Matrix Description of NMR BCMB/CHEM 8190

Existence and Uniqueness Results for Random Impulsive Integro-Differential Equation

The Analysis of the Thickness-predictive Model Based on the SVM Xiu-ming Zhao1,a,Yan Wang2,band Zhimin Bi3,c

Cubic Bezier Homotopy Function for Solving Exponential Equations

Survival Analysis and Reliability. A Note on the Mean Residual Life Function of a Parallel System

Dynamic Team Decision Theory. EECS 558 Project Shrutivandana Sharma and David Shuman December 10, 2005

Asymmetry and Leverage in Conditional Volatility Models*

Math 128b Project. Jude Yuen

Variance Stabilizing Power Transformation for Time Series

Stochastic Maxwell Equations in Photonic Crystal Modeling and Simulations

Notes on the stability of dynamic systems and the use of Eigen Values.

Solving Parabolic Partial Delay Differential. Equations Using The Explicit Method And Higher. Order Differences

Machine Learning 2nd Edition

Bayesian Inference of the GARCH model with Rational Errors

Density Matrix Description of NMR BCMB/CHEM 8190

( ) () we define the interaction representation by the unitary transformation () = ()

Epistemic Game Theory: Online Appendix

On the numerical treatment ofthenonlinear partial differentialequation of fractional order

Linear Response Theory: The connection between QFT and experiments

CS286.2 Lecture 14: Quantum de Finetti Theorems II

Data Collection Definitions of Variables - Conceptualize vs Operationalize Sample Selection Criteria Source of Data Consistency of Data

Degrees of Freedom. Spherical (ball & socket) 3 (3 rotation) Two-Angle (universal) 2 (2 rotation)

Implementation of Quantized State Systems in MATLAB/Simulink

Mechanics Physics 151

Robustness of DEWMA versus EWMA Control Charts to Non-Normal Processes

Motion of Wavepackets in Non-Hermitian. Quantum Mechanics

On Pfaff s solution of the Pfaff problem

On Convergence Rate of Concave-Convex Procedure

Sensor Scheduling for Multiple Parameters Estimation Under Energy Constraint

The Performance of Optimum Response Surface Methodology Based on MM-Estimator

Mechanics Physics 151

Solution in semi infinite diffusion couples (error function analysis)

A Paper presentation on. Department of Hydrology, Indian Institute of Technology, Roorkee

Lecture 6: Learning for Control (Generalised Linear Regression)

Stochastic Reliability Analysis of Two Identical Cold Standby Units with Geometric Failure & Repair Rates

Solution for singularly perturbed problems via cubic spline in tension

The Finite Element Method for the Analysis of Non-Linear and Dynamic Systems

Bayes rule for a classification problem INF Discriminant functions for the normal density. Euclidean distance. Mahalanobis distance

Multivariate Ratio Estimator of the Population Total under Stratified Random Sampling

CHAPTER 5: MULTIVARIATE METHODS

. The geometric multiplicity is dim[ker( λi. number of linearly independent eigenvectors associated with this eigenvalue.

EEL 6266 Power System Operation and Control. Chapter 5 Unit Commitment

EFFICIENCY IMPROVEMENTS FOR PRICING AMERICAN OPTIONS WITH A STOCHASTIC MESH: PARALLEL IMPLEMENTATION 1

Motion in Two Dimensions

Lecture 18: The Laplace Transform (See Sections and 14.7 in Boas)

Li An-Ping. Beijing , P.R.China

Sampling Procedure of the Sum of two Binary Markov Process Realizations

Method of upper lower solutions for nonlinear system of fractional differential equations and applications

Scattering at an Interface: Oblique Incidence

WiH Wei He

Improvement in Estimating Population Mean using Two Auxiliary Variables in Two-Phase Sampling

Volatility Interpolation

Transcription:

J. Sa. Appl. Pro., o., -7 SP Journal of Sascs Applcaons & Proaly --- An Inernaonal Journal @ SP aural Scences Pulsn Cor. On Consrucon of Odd-fraconal Facoral Desns Ike Basl Onukou Deparmen of Sascs, Unversy of Uyo, Akwa Iom Sae, era Emal Address: keaslonukou@yaoo.com Receved: Dec., ; Revsed Fe., ; Acceped Fe. 8, Pulsed onlne: Aprl Asrac: Fraconal desns nvolve selecon from a ven se of expermenal reamens as suse of reamens o makeup a specfed desn measure a as suc sascal properes as alance, relave effcency, D-opmaly ec. For decades sascans ave reled on e use Defnn Conracs DC, and Lan Squares LS o consruc fraconal facoral desns. Bu ese meods are sown o ave very lmed rane of applcaons and somemes produce desns a are snular. Ts paper nroduces e meod of Concenrc Balls for consrucn non-snular fraconal desns. Eac all consss of reamens a are of equal dsance from e cener and usn a se of rules for selecn reamens from a all e meod yelds a small se of admssle desns. Te es memer of s admssle se s e desred desn:{bes n e sense of maxmzn e deermnan of e normalzed nformaon marx or maxmzn e relave effcency of e facoral effecs.}umercal examples sow a e meod covers every rane of expermenal desn condons and can produce fraconal desns a are D-opmal. Keywords: Odd-fracon, concenrc alls, relave effcency Inroducon Consrucon of fraconal facoral desns s a opc a s exensvely reaed n mos sandard exs on desn of expermens; see, e.. Cocran and Cox 957, Anderson and Mclean 974. From n-ndependen, non-socasc varales, were e varae, x appears a s -levels, we e s s s n reamens and consder ree knds of reamen spaces : Te unform or symmerc form; s x, x,, xn; x,,, s, s s sn, Te non-unform or asymmerc ype; A x,, xn; s s. for a leas one par of, Te Irreular ype; ; e.. x, x,, x ; x,,, s, R n Oer eomerc forms can also occur.e. x, x,, xn ; x,,,, n s a produc of connuous nervals; owever, e coverae of s repor does no nclude connuous nervals. As saed earler, e prolem of neres ere s o consruc an -pon desn p,.e. an en e numer of parameers n e response s s s n fraconal facoral desn, p funcon f x. Te fracon s consdered an odd-fracon f s no dvsle y any s, oerwse s a reular fracon. For decades, e pracce as een o consruc fraconal facoral desns usn eer Lan Squares LS or Defnn Conrass DC; see, e.. Anderson and Mclean. Two prolems can arse from s approac: a Te DC and LS meods are napplcale, as n. 7

J. Sa. Appl. Pro., o., -7 Te meods can produce snular or near snular desns as sown n ale..., even wen e relave effcency of e desn s consdered ood. Ts paper nroduces e Concenrc Balls meod of consrucon a as a wde rane of applcaons and can produce an admssle se of equvalen desns, leavn e scens o make a coce. Te meod proceeds as follows: Arrane e suppor pons no H roups or alls, so a suppor pons a are of e same dsance from e cener are n one all. Tus e all, x, x,, xn conans n suppor pons, =,,..., H, xk s an n-componen vecor, k =,,...,n, were, d x xk s e dsance from e cener, and d d d. Paron k H no su-roups accordn o e numer of neave sns and zeros appearn. a e suppor pon xk; see secon ree of s paper. Apply e selecon rules; see secon wo o uld up e requred desn. Tese rules yeld a small se of admssle desns wose deermnans and relave effcences can e easly compared. Applcaon of e dea of roupn of reamens owards consrucon of D-opmal exac desns ave een employed y Onukou and Iwundu 7; and for D-opmals desns for -level facoral models and auoreressve error y Ye and Huan5. Consrucon and analyss of fraconal facoral desns on a wder plaform as een consdered y Guns and Mason 9.A rane of ecnques for consrucon of asymmerc fraconal facorals as well as condons for nonexsence of e desns ave een ven y Dey and Raul 999. A way as offered y Oludua and Madukafe 9 for serean fraconal facoral desns on e ass of er D-opmal and loss of nformaon values. As lon as neres n a facoral expermen s resrced o a lmed numer of parameers facoral effecs researc n fraconal desns wll connue o flours. In wa follows, e asc alera for e ecnque s dscussed n secon wo, wle numercal llusraons are ven n secon ree.. Alerac Bass Te expermenal space wll e represened y e rple, F, ; x, x,, x ; x,, s,,, n s a connuous, compac, merc space of rals, x x n Fx f x; x s a se of connuous, dfferenale funcons. x x; x s a se of connuous, non-neave error funcons. Eac se of e rple s consdered fne and oeer ey form a ass for n-dep sudy of e sujec of desn of expermens; see, e.. Pazman 987, Aknson and Donev 99, Onukou 997. Le f x e a frs-order neracve funcon defned y. f x e = x j s an p n n p exended desn marx; e p parameers comprsn e lnear and neracve erms,, s an lock ncdence marx; da k, k,, k ; k j en e sze of e j lock, s a p-parameer vecor of reamen effecs e s an -componen vecor of random error Te deermnan of e nformaon marx n. equals, k j de de I R; R j s e marx of loss of nformaon..

Pranes Kumar: Sascal Dependence: Copula Funcons A eomerc meann of loss of nformaon as cos. of e anle of nclnaon of a reamen effec on e locks as een reaed y Onukou. ow, for an -pon desn n one lock =, R rr ; r r, r,, r ; r,,,, p.; r s e loss of nformaon on e reamen effec. Hence, e eomerc mean, p. r r p p ves a measure of e overall effcency of e desn relave o a complee lock desn. oce a for r, r. s maxmzed wen e desn s alanced; e. wen r r r. Bu r. does no p ake no accoun e deermnan, de, and erefore can ake non-zero values for snular desns. Bu y ncludn e deermnan, we e e creron for comparn desns:.4 d m r ; m de / To maxmze.4 e follown selecon rules are o e appled wen makn-up e desn measure : max x j mn xj mn xj xj, j,,, p, j j We recall a x s e exended reamen marx. j Realsn a e numer of suppor pons n n s n p n n ; en, for e response funcon.,e opmal -pon desn s consruced from no e case for a complee quadrac funcon, n.5 f x a a x a x x a x e n only. Bu, s s Te sarn pon of e procedure s dependen on e relave values of and p. Wen = p, e procedure oans e desn as follows: a A e nal sep ake p n suppor pons from and e res of p + n + from, y applcaon e rules n.4 aove. Ts yelds a se of r admssle desns S A.,, r m max m ; S A Compue c A e k sep ake p + k n suppor pons from k k m k max m ; SA d Sop, f k k k m m m and e res from, and compue For >> p, e aove sequence can en y akn p or p+ pons from. I seems a y properly relan e numer of pons o e aken from o e rao p/, sould e possle o develop a nonerave procedure for consrucn opmal fraconal facoral desns for quadrac response funcons.. umercal Examples Gven are rvarae frs-order neracve response surface,. f x, x, x a ax ax ax axx axx axx e and a cuc surface, x, x, x ; x,,,,,.

4 Pranes Kumar: Sascal Dependence: Copula Funcons Usn eac of e ree meods, we consder e consrucon of wo fraconal facoral desns: 9 a, for a reular fracon, and 7, for an odd-fracon. 7 Te reamen ale s ven y. Tale of Treamens For x,, x,, x,, cuc surface Usn for nsance xx x as defnn conras; see, e.. Anderson and Mclean, for deals, e DC meod produces e desn 9 DC, wereas e Lan square meod ves 9 LS Te roups and suroups requred for e meod are: - - - - - + - + - + - - + + - + - + - + + + + + - - - - - - - + + - - + + - - + + - + + + + + + 4 Applcaon of e rules under.4 ves wo equvalen desns, Te deermnans of ese desns are repored n Tale. 4 9, 9.. Deermnans, De. And Relave Effcences, RE. For Tree Meods of Consrucn Fraconal Desns for Frs-Order Ineracve Funcons - - - + + +

Pranes Kumar: Sascal Dependence: Copula Funcons 5 Seral umer Fraconal Desn METHOD OF COSTRUCTIO DC LS De. RE. De. RE. De. RE. 9.546.98.546.98 4.4 7 x E-5 x E-5 x E- A A A A 69.95 7 x E- 4 A A.9. 7.757 5 x E- x E- 4 7 A A A A 9.5486 5 x E-.99985.9988.99875.9999 A means o Applcale. Smlarly, for e odd-fracon e meod produces wo equvalen desns: 7, Bo e DC and LS meods are napplcale n s case of odd-fracon. If e space of rals s non-unform asymmerc r-varae surface, x, x, x; x, x,, x,,,,, We consder e consrucon of wo desns: c 4 5 for reular fracon, and d 7 5 for odd-fracon. Jus as n ale., a correspondn reamen ale s se-up and snce 4 s dvsle y, e Lan- Square meod can e appled. Te frs sep s o se-up a Paral Lan Square PLS, L usn e leers a and. ex, supermpose L on e frs ac of 5 reamens were x, and en on e oer ac of 5 reamen a x, a L a a a a a a a Fnally, e reamens a concde w e leer are rouped oeer o form e desn: 4 PLS, de. =.9 x E - On e conrary e meod produces an opmal desns;

Pranes Kumar: Sascal Dependence: Copula Funcons 6 4 w de. = 7.759 x E - Te consrucon of 5 7 also ves wo equvalen desns: 7 and 7 w de. = 9.5486 x E - ; oce a all e desns for e frs-order neracve funcons are consruced from e frs all for all values of. Ts owever s no e case for quadrac response funcon defned n.5. Apparenly, wo concenrc alls are requred o consruc a desn for quadrac funcons. For example, wo equvalen desns are oaned y e meod for a ; 7 namely, and w 4588.. m x E -4. Smlarly, for e asymmerc surface, e meod ves wo equvalen opmal desns for 5; namely, ; 4 4 and 4 4,, s ven aove. Summary and Concluson Te paper as sown a DC meod can e used o consruc fraconal facoral desns only wen e facor levels are unform and even a s, e relave effcency of s meod s comparavely nferor. On e oer and, e Lan square meod can e appled o for unform and non-unform levels provded only a e fracon s reular. Of e ree meods s only e

Pranes Kumar: Sascal Dependence: Copula Funcons 7 meod a can consruc odd-fraconal facoral desns; desns w e es level of effcency; desns a are requred o e alanced n one replcaon as well as desns a are D-opmal. Acknowledemen We are raeful o e Revewer and e Edor-n-Cef for e references numers and 4. References []. Anderson, V.L. and Mclean, R.A. 974: Desn of Expermens: A Realsc Approac, Marcel Dekker. []. Cocran, W.G. and Cox, G.M. 957: Expermenal Desns, nd Edon, J.Wley []. Dey, Aloke and Mukerjee, Raul 999: Fraconal Facoral Plans; Jon Wley and Sons, ew York. [4]. Guns, Rcard, F. and Mason, Roer, L 9: Fraconal Facoral Desn; Wley Inerdsplnary Revews: Compuaonal Sascs, Vol., Issue, Pae 4 44 [5]. Oludua, A.V. and Madukafe, M.S. 9: D-Opmaly and D L -Opmaly crera for Incomplee Block Desns; Gloal Journal of Maemacal Scences, Vol. 8, o. Pae 7 5. [6]. Onukou, I.B. 997: Foundaons of Opmal Exploraon of Response Surfaces, Epraa Press, sukka, era. [7]. Onukou, I.B. and Iwundu, M.P. 7: A Comnaoral Procedure for Consrucn D-Opmal Exac Desns; Sasca Rvsa Vol.67 Pae 45 4. [8]. Pazman, A. 987: Foundaon of Opmum Expermenal Desn, Redel Pulsn Company. [9]. Ye, Hon-Gwa and Huan, Mon-a Lo 5: On Exac D-Opmal Desns w wo-level facors and n auocorrelaed