Statistics Assessing Normality Gary W. Oehlert School of Statistics 313B Ford Hall

Similar documents
Jonathan Turner Exam 2-10/28/03

1 Finite Automata and Regular Expressions

Relation between Fourier Series and Transform

Week 06 Discussion Suppose a discrete random variable X has the following probability distribution: f ( 0 ) = 8

3.4 Repeated Roots; Reduction of Order

Math 266, Practice Midterm Exam 2

Laplace Transform. National Chiao Tung University Chun-Jen Tsai 10/19/2011

Advanced Engineering Mathematics, K.A. Stroud, Dexter J. Booth Engineering Mathematics, H.K. Dass Higher Engineering Mathematics, Dr. B.S.

Walk Like a Mathematician Learning Task:

EE Control Systems LECTURE 11

1 Introduction to Modulo 7 Arithmetic

Erlkönig. t t.! t t. t t t tj "tt. tj t tj ttt!t t. e t Jt e t t t e t Jt

Fourier Series and Parseval s Relation Çağatay Candan Dec. 22, 2013

Instructions for Section 1

The Laplace Transform

Library Support. Netlist Conditioning. Observe Point Assessment. Vector Generation/Simulation. Vector Compression. Vector Writing

graph of unit step function t

Single Correct Type. cos z + k, then the value of k equals. dx = 2 dz. (a) 1 (b) 0 (c)1 (d) 2 (code-v2t3paq10) l (c) ( l ) x.

Introduction to Laplace Transforms October 25, 2017

Lecture 1: Numerical Integration The Trapezoidal and Simpson s Rule

The Mathematics of Harmonic Oscillators

Revisiting what you have learned in Advanced Mathematical Analysis

Section 3: Antiderivatives of Formulas

Mathcad Lecture #4 In-class Worksheet Vectors and Matrices 1 (Basics)

A Tutorial of The Context Tree Weighting Method: Basic Properties

Floating Point Number System -(1.3)

Floating Point Number System -(1.3)

Ch 1.2: Solutions of Some Differential Equations

Wave Phenomena Physics 15c

Pupil / Class Record We can assume a word has been learned when it has been either tested or used correctly at least three times.

FL/VAL ~RA1::1. Professor INTERVI of. Professor It Fr recru. sor Social,, first of all, was. Sys SDC? Yes, as a. was a. assumee.

2. The Laplace Transform

where: u: input y: output x: state vector A, B, C, D are const matrices

What do you know? Listen and find. Listen and circle. Listen and chant. Listen and say. Lesson 1. sheep. horse

Shortest Paths. CSE 421 Algorithms. Bottleneck Shortest Path. Negative Cost Edge Preview. Compute the bottleneck shortest paths

Engine Thrust. From momentum conservation

Midterm exam 2, April 7, 2009 (solutions)

Chem 104A, Fall 2016, Midterm 1 Key

16.512, Rocket Propulsion Prof. Manuel Martinez-Sanchez Lecture 3: Ideal Nozzle Fluid Mechanics

ECE COMBINATIONAL BUILDING BLOCKS - INVEST 13 DECODERS AND ENCODERS

RUTH. land_of_israel: the *country *which God gave to his people in the *Old_Testament. [*map # 2]

Trigonometric Formula

Final Exam : Solutions

Instructors Solution for Assignment 3 Chapter 3: Time Domain Analysis of LTIC Systems

Lecture 11 Waves in Periodic Potentials Today: Questions you should be able to address after today s lecture:

Option markets and the stochastic behavior of commodity prices 1

The Procedure Abstraction Part II: Symbol Tables and Activation Records

Boyce/DiPrima 9 th ed, Ch 2.1: Linear Equations; Method of Integrating Factors

Lecture contents. Bloch theorem k-vector Brillouin zone Almost free-electron model Bands Effective mass Holes. NNSE 508 EM Lecture #9

DEPARTMENT OF ELECTRICAL &ELECTRONICS ENGINEERING SIGNALS AND SYSTEMS. Assoc. Prof. Dr. Burak Kelleci. Spring 2018

Applications of these ideas. CS514: Intermediate Course in Operating Systems. Problem: Pictorial version. 2PC is a good match! Issues?

Overview. Splay trees. Balanced binary search trees. Inge Li Gørtz. Self-adjusting BST (Sleator-Tarjan 1983).

CSE 373: More on graphs; DFS and BFS. Michael Lee Wednesday, Feb 14, 2018

Chapter 5 The Laplace Transform. x(t) input y(t) output Dynamic System

The model proposed by Vasicek in 1977 is a yield-based one-factor equilibrium model given by the dynamic

Multi-Section Coupled Line Couplers

a b c cat CAT A B C Aa Bb Cc cat cat Lesson 1 (Part 1) Verbal lesson: Capital Letters Make The Same Sound Lesson 1 (Part 1) continued...

T h e C S E T I P r o j e c t

AR(1) Process. The first-order autoregressive process, AR(1) is. where e t is WN(0, σ 2 )

Elementary Differential Equations and Boundary Value Problems

A Study of the Solutions of the Lotka Volterra. Prey Predator System Using Perturbation. Technique

Ma/CS 6a Class 15: Flows and Bipartite Graphs

COMP108 Algorithmic Foundations

Blended Level 1 and Level 2 Sample Lesson Plans

TOPIC 5: INTEGRATION

CSE 421 Algorithms. Warmup. Dijkstra s Algorithm. Single Source Shortest Path Problem. Construct Shortest Path Tree from s

Blended Level 1 and Level 2 Sample Lesson Plans

(A) 1 (B) 1 + (sin 1) (C) 1 (sin 1) (D) (sin 1) 1 (C) and g be the inverse of f. Then the value of g'(0) is. (C) a. dx (a > 0) is

Partial Fraction Expansion

Payroll Direct Deposit

Present state Next state Q + M N

NEW FLOODWAY (CLOMR) TE TE PIN: GREENS OF ROCK HILL, LLC DB: 12209, PG: ' S67 46'18"E APPROX. FLOODWAY NEW BASE FLOOD (CLOMR)

3) Use the average steady-state equation to determine the dose. Note that only 100 mg tablets of aminophylline are available here.

Bipartite Matching. Matching. Bipartite Matching. Maxflow Formulation

Generalized Half Linear Canonical Transform And Its Properties

CS 541 Algorithms and Programs. Exam 2 Solutions. Jonathan Turner 11/8/01

The Theory of Small Reflections

The Transfer Function. The Transfer Function. The Transfer Function. The Transfer Function. The Transfer Function. The Transfer Function

Using the Printable Sticker Function. Using the Edit Screen. Computer. Tablet. ScanNCutCanvas

Chapter 2 The Derivative Business Calculus 99

P a g e 5 1 of R e p o r t P B 4 / 0 9

Chapter4 Time Domain Analysis of Control System

T HE 1017TH MEETING OF THE BRODIE CLUB The 1017th Meeting of the Brodie Club was held at 7:30 pm on January 15, 2008 in the R amsay Wright Laboratorie

Weighted graphs -- reminder. Data Structures LECTURE 15. Shortest paths algorithms. Example: weighted graph. Two basic properties of shortest paths

Chapter 5 Transient Analysis

Fourier. Continuous time. Review. with period T, x t. Inverse Fourier F Transform. x t. Transform. j t

Planar Upward Drawings

NEWBERRY FOREST MGT UNIT Stand Level Information Compartment: 10 Entry Year: 2001

CHARACTERIZATION FROM EXPONENTIATED GAMMA DISTRIBUTION BASED ON RECORD VALUES

INTERQUARTILE RANGE. I can calculate variabilityinterquartile Range and Mean. Absolute Deviation

Derivation of the differential equation of motion

Errata for Second Edition, First Printing

On the Existence and uniqueness for solution of system Fractional Differential Equations

Factors Success op Ten Critical T the exactly what wonder may you referenced, being questions different the all With success critical ten top the of l

HIGHER ORDER DIFFERENTIAL EQUATIONS

(2) If we multiplied a row of B by λ, then the value is also multiplied by λ(here lambda could be 0). namely

b. How many ternary words of length 23 with eight 0 s, nine 1 s and six 2 s?

x, x, e are not periodic. Properties of periodic function: 1. For any integer n,

NUCON NRNON CONRNC ON CURRN RN N CHNOOGY, 011 oo uul o w ul x ol volv y y oll. y ov,., - o lo ll vy ul o Mo l u v ul (G) v Gl vlu oll. u 3- [11]. 000

SOLVED EXAMPLES. be the foci of an ellipse with eccentricity e. For any point P on the ellipse, prove that. tan

Transcription:

Siic 504 0. Aing Normliy Gry W. Ohlr School of Siic 33B For Hll 6-65-557 gry@.umn.u Mny procur um normliy. Som procur fll pr if h rn norml, whr ohr cn k lo of bu n kp going. In ihr c, i nic o know how clo o norml h r. Aing normliy i n xmpl of goon-of-fi problm. Goon-of-fi i ifficul problm. Thr r mny pproch. W will mphiz grphicl pproch n h numricl follow-on o h grph. Sr wih h univri c. Q-Q plo. Q-Q i hor for qunil-qunil. Whn ing normliy, h r lo known norml probbiliy plo n (loclly) rnki plo. Mk pir, whr i h h qunil of om horicl iribuion, n i h h qunil of h. Thn mk plo of h pir. If h com from h horicl iribuion, h poin on h plo houl fll on lin wih lop n inrcp 0. Plo will b linr if qunil r cl qunil plu conn:. +,.-/ Uully w choo lvl omhing lik or! #"%$ or &' ( *) for " " ", o h h qunil r ju h orr 035476# 80:9;4<6 6# >0@?A4 5 "=" ". For coninuou horicl iribuion wih cumuliv B, h horicl qunil r ju BDC =E. I gnrlly mk lil iffrnc which vrion of AE w u. Cm> r("",r,r,h,h,u,u) R from fil "/crom/t-8.dat" Column v REAL vcor r Column v REAL vcor r Column 3 v REAL vcor h Column 4 v REAL vcor h Column 5 v REAL vcor u Column 6 v REAL vcor u Cm> q <- invnor((run(5)-.5)/5) Cm> q <- invnor(run(5)/6) Cm> q3 <- invnor((run(5)-3/8)/(5.5)) Cm> cor(hconc(q,q,q3)) (,) 0.999 0.9999 (,) 0.999 0.9996

(3,) 0.9999 0.9996 Cm> cor(hconc(q3,rnki(q3))) (,) (,) Cm> or <- gr(r) Cm> plo(q3,r[or], xlb:"norml qunil", ylb:"or ", il:"dominn riu"). Dominn riu o r 0.9 0.8 0.7 0.6 0.5 - -.5 - -0.5 0 0.5.5 norml qunil Cm> plo(rnki(h),h, xlb:"norml qunil", ylb:"or ",il:"humru")

Humru. o r.8.6.4 - -.5 - -0.5 0 0.5.5 norml qunil Cm> yll <- hconc(r,r,h,h,u,u) Cm> yll <- or(yll) Cm> chplo(q3,yll,lin:t, xlb:"norml qunil", ylb:"or ",il:"all ") o r All 3. 3 3 34 4.8.6 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 4 4 4 3 3 3.4 3 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4. 3 0.8 0.6 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 56 56 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 - -.5 - -0.5 0 0.5.5 norml qunil A of normliy, conir h corrlion of h norml qunil n h qunil. Lrg vlu of h corrlion r conin wih normliy; mll vlu r inconin. Cm> cor(q3,yll)[,] (,) 0.948 0.970 0.983 0.99 (,6) 0.984 0.995 3

How mll i oo mll? No impl nwr, o imul h iribuion. Cm> r <- rp(0,0000) Cm> for(i,run(0000)) x <- or(rnorm(5)) r[i] <- cor(q3,x)[];; Cm> min(r) () 0.8655 Cm> mx(r) () 0.99734 Cm> hi(r,run(.85,,.005)) 50 Hiogrm of r wih ol r 40 D n i y 30 0 0 0 0.86 0.88 0.9 0.9 0.94 0.96 0.98 r Cm> obr <- cor(q3,yll)[,-];obr (,) 0.948 0.970 0.983 0.99 (,6) 0.984 0.995 Cm> um(r<obr)/0000 (,) 0.03 0.44 0.47 0.86 0.503 (,6) 0.984 Th r h imul p-vlu for ing normliy for h ix vribl. I m no complly hppy wih mny of normliy, incluing hi on. Whn i mll, hy hv low powr. 4

- ' " Whn i big, hy cn c viion from normliy h migh b f o ignor. Thy n o b u in conjucion wih jugmn n ohr mho, uch mn of h QQ plo. Wh o w o for mulivri norml? Uul pproch i o if propri known o hol for MVN hol for h. Norml mrginl, norml coniionl, norml linr combinion, linr rgrion of ch vribl on ohr vribl, conn coniionl vrinc, chi-qur inc, c. Whn w normliy for ch vribl, w r ing mrginl normliy. Rcll h p-vlu: Cm> um(r<obr)/0000 (,) 0.03 0.44 0.47 0.86 0.503 (,6) 0.984 Shoul w rjc normliy bcu vribl on h p-vlu of.03? No ncrily. W hv mulipl ing iuion (lo cll mulipl comprion or imulnou infrnc). In hi problm w hv null hypoh H E (normliy of vribl ) n n ovrll null hypohi H which i ru if ll h H E r ru. If w ch H E lvl, hn 6 6 rjc H n o b clor o for mll, bu high corrlion mk clor o. Th Bonfrroni jumn y o rjc H if h mll iniviul p-vlu i l hn <, or quivlnly if im h mll p-vlu i l hn, whr i h numbr of. Th Bonfrroniz h. In our xmpl " ',( i no vry mll, n w woul no k h mrginl rul rong vinc gin mulivri normliy. On common rcommnion i o chck h normliy of, whr i h mrix of ignvcor for, h vrinc mrix of. Thi ro h o x of h llip n o norml ing mrginlly own ch xi. Th wor llip i in quo bov, bcu w migh no hv llipicl poin clou for nonnorml. Cm> V <- b(yll,covr:t) Cm> U <- ign(v)$vcor Cm> w <- yll%*%u; w <- or(w) Cm> obrw <- cor(q3,w)[,-];obrw (,) 0.990 0.968 0.985 0.970 0.987 (,6) 0.986 Cm> um(r<obrw)/0000 (,) 0.808 0. 0.57 0.49 0.650 (,6) 0.605 5

E C If N <, hn & If i lrg, hn 9 + E!C E houl lo b pproximly 9 9. U Q-Q plo wih E n horizonl vlu h prcn poin of 9. Th qur roo of h vlu will omim work br for mll. Cm> <- icomp(yll); () 8.866 7.459 0.848 3.59 4.374 (6) 0.83 3.569 3.08 3.54.600 ().570 9.40.897.74 5.59 (6) 7.5 6.93.560 7.74 7.70 ().057 7.345 3.9 3.607 7.4 Cm> <- or() Cm> q <- invchi((run(5)-.5)/5,6) Cm> plo(q,, xlb:"chi-qur wih 6 f qunil", ylb:"qur inc", il:"qq inc plo for bon ", how:f) Cm> lin(vcor(0,5),vcor(0,5), how:f) Cm> howplo(xmin:0,ymin:0) 9 4 QQ inc plo for bon q u r i n c 0 8 6 4 0 0 4 6 8 0 4 chi-qur wih 6 f qunil 6

Cm> plo(qˆ.5,ˆ.5, xlb:"chi wih 6 f qunil", ylb:"inc", il:"qq inc plo for bon ", how:f) Cm> lin(vcor(0,4),vcor(0,4), how:f) Cm> howplo(xmin:0,ymin:0) 3.5 3 QQ inc plo for bon i n c.5.5 0.5 0 0 0.5.5.5 3 3.5 chi wih 6 f qunil W migh wn o buil b on hi plo, bu w cn u corrlion. W rlly wn i o b on h lin wih inrcp 0 n lop. Try viion from h lin, or wigh viion from h lin. For xmpl, (-q)ˆ or (-q)ˆ/q. I upc h lrgr orr iic hv mor vrinc, o I my wn o ownwigh hm. Myb iviing by q (mor or l h xpc vlu) will work. Smpl bunch of chiqur, orr hm, n compu vrinc. Cm> D <- mrix(invchi(runi(5000),6),5) Cm> D <- or(d) Cm> D <- D Cm> b(d,vr:t) () 0.86 0.39 0.340 0.375 0.40 (6) 0.44 0.448 0.47 0.498 0.549 () 0.58 0.6 0.654 0.700 0.785 (6) 0.833 0.908.086.90.36 ().658.044.886 4.467 0.4 7

Cm> b(d,vr:t)/q () 0.5 0.86 0.54 0.46 0.38 (6) 0.3 0.7 0.3 0. 0.4 () 0.3 0. 0. 0.3 0.30 (6) 0.30 0.34 0.50 0.54 0.65 () 0.87 0. 0.7 0.370 0.673 Cm> b(d,vr:t)/qˆ () 0. 0.06 0.070 0.057 0.047 (6) 0.04 0.036 0.03 0.09 0.08 () 0.06 0.04 0.03 0.0 0.0 (6) 0.00 0.00 0.0 0.00 0.00 () 0.0 0.0 0.05 0.03 0.045 Cm> D <- D Cm> ou <- um( (D-q)ˆ/q) Cm> ou <- vcor(ou) Cm> um( (-q)ˆ/q) ().009 Cm> lngh(ou) () 000 Cm> um(ou >.009)/000 () 0.8 Thi ming roun wih q cling fcor i no h b w cn o. W im h vrinc mrix of h orr iic uring our imulion. W cn mk br uing ho vrinc. 8