This preview has intentionally blurred sections. Sign up to view the full version.
View Full DocumentThis preview has intentionally blurred sections. Sign up to view the full version.
View Full DocumentThis preview has intentionally blurred sections. Sign up to view the full version.
View Full DocumentThis preview has intentionally blurred sections. Sign up to view the full version.
View Full DocumentThis preview has intentionally blurred sections. Sign up to view the full version.
View Full Document
Unformatted text preview: %NUS
\li’fl)’ :1‘1'3'213LZ” V‘ " Chagter 0 Basic Prerequisite Knowledge ST3‘I 31 Regression Analysis > 0.1 1. Distributions  Normal Distribution, N01, 02) /  Chisquare Distribution, x2 (n)
— Let U x ZZ. IfZ ~ N(O, 1), then (mummy Cid ~ 1' 1 I U ~ x20)
Definition \
— Let Wm X1 + + X”. IfX, ~ x2(1],i= 1, ..., n,
independently, then
W ~ XZW
Deﬁnition9=
LEV Zi;“'Zh N Nb ‘
ST3131 Re ression Anal sis
g y W: 212+ 221*“ W
' t—Distribution, t(mf day/€94 9% ham“ — IfZ w NU), 1) and V N X2072) independently, then  FDistribution, F (m, n)
— If V ~ szz) and W ~ x201) independently, then F V/m F(
m m w my:
W/n 7‘ I. Wraneécr 9
pmmm' i
ST3131 Regression Analysis \ / 0.3 can not Ohqu 03rde 2. Confidence Interval A
o If § is a point estimate of 9, which follows a
normal. or an approximate normal distribution, :5 6‘9 JV M O] ‘ >
0 then a 100(1 — a)% confidence interval for G is ‘5 . A z
giVen by Wt (V is anknbwn,9§b‘mm JOYEQM‘DD
(9 i t m 5.8.(é) v“  N l 2.
9’1 (Rf/2 (y)
0‘
where s.e.(é) is the standard error (Le. the + rgQié) f and é’Q are mcingMient
estimated standard deviation) of § [0: r V v —~ ‘ x
t) (9 “g m ‘i: i V) p
a 99(5). .—/ — 4‘ M
\ ‘ ‘ kg
(,thfrden Le Miami ( / ST3131 Regression Analysis 04 gamus
\gjwy 3. Review of Matrices 3.1 Notation  A = ((117),: 1, m: 1’ q denotes a p X C] matrlx. €11  _C_l_ = 5 denotes a p—dimensional vector.
at?  A’ = (aﬁ) = transpose ofA.  110 = p X [9 identity matrix.  lp = pdimensional vector of 1’s. ST3131 Regression Analysis 0.5 3. Review of Matrices 3.1 Notation
IfA is a p X p matrix, then
 M] x determinam OfA (or det(A)), 0 A4 : inverse ofA, (i.e.AA‘1 = A‘lA = I .)
PW a WM 2x), mah‘rx ; (0; iiiimitzzﬂi “ii ST3131RegressionAnalysis T ‘ a b .6
dece'rminant <4: ( C d8 NUS
‘ 333;: 3.2 Exgectation X1 Y1
 Let g m 5 and}: m 5 be random vectors
X39 Y}?
5 (X1) 50/1)
 Then 5(5) m 5 and EQ) m 3
503(3)) E(Yp)  HA is a q X p matrix of constants and b is a q X 1
vector of constants, then EMQQ+§FAHXQ+Q ST3131 Regression Analysis 0.7 N us
\Lftg' 3.3 Covariance Matrix  Covariance matrix is defined as 601201,!) —— E [({1— E(£)) (X — E(z))'] .
E (Xi — maim — 1509)) Eiwi—Ecxiiﬁmwmi (5.014% 3/1)). . i=1,...1p;j:1’...’p ST3131 Regression Analysis 0.8 (J4; ‘\ 3.3 Covariance Matrix (Continued)  When X = )_(, Cal/(X, )_Q is called the variance—
Covariance matrix of )_( (or dispersion matrix) and is
denoted by VQQ. W) 3 60mg.) Warm) Cay(X3,X2) maﬁa?)
m Cov(X2,X1) Von/(X2) Cev(X2,Xp)
VartXp) ST3131 Regression Analysis 0.9 3.3 Covariance Matrix (Continued) 0 Note: WA) is asymmetric matrix. (Iv/(>521) I\s 043m o {rmmétrtc Matrix  HA is a p X 19 matrix of constants and b is a p X 1
vector of constants, then Wt gm z_;) :A VQQ A’.
o In particular, WA 2Q 2 A VQQ A’.  Note: Adding a constant vector b to the random
vector AX does not change the variance of A)_(. ST3131 Regression Analysis 010 3.4 Some grogerties of matrices
1. AB i BA Z. {A’}’ m 14% and {A}??? m 32%” 3. AB “1 x 31144 providin A‘1 and B“1 exist
8 ST3131 Regression Analysis 0“ ...
View
Full Document
 Spring '10
 LI

Click to edit the document details