Académique Documents
Professionnel Documents
Culture Documents
Sabemos que:
σ 2
X ~ N µ, ⇒ Z = X − µ ~ N ( 0,1 )
n σ/ n
Então:
X−µ
P −z(α / 2 )< < z(α / 2 ) = 1 − α
σ/ n
σ σ
P µ − z ( α / 2 )⋅ < X < µ + z ( α / 2 )⋅ =1− α
n n
ou como:
σ σ
P X − z ( α / 2 )⋅ < µ < X + z ( α / 2 )⋅ =1− α
n n
σ σ
X − z ( α / 2 )⋅ , X + z ( α / 2 )⋅
n n
NOTA:
α1 + α 2 = α
os intervalos
σ σ
X − z ( α )⋅ , X + z ( α )⋅
n
1 2
n
σ 2
X ~ N µ, ⇒ Z = X − µ ~ N ( 0,1 )
n σ/ n
⋅ ∑ ( Xi − X )
1 n 2
S=
n − 1 i =1
S≈σ (constante)
e portanto:
X−µ X−µ
Z= ≈ ~ N ( 0,1 )
S/ n σ/ n
S
X − z ( α / 2 ) ⋅
S
, X + z ( α / 2 ) ⋅
n n
S≈σ (constante)
X−µ X −µ
≈
S/ n σ/ n
212
X −µ
S/ n
Notemos que:
X −µ
X−µ N ( 0,1 )
= σ/ n ~
S/ n S/ σ χ 2n −1
n −1
X−µ
e como e S / σ são v.a. independentes, resulta da
σ/ n
definição da distribuição t de Student que:
X−µ
~ t n −1
S/ n
S
X − t n −1 ( α / 2 ) ⋅ n , X + t n −1 ( α / 2 ) ⋅ n
S
213
INTERVALOS DE
CONFIANÇA PARA A
PROPORÇÃO BINOMIAL ( P = Y )
n
p ⋅ (1 − p )
P = Y ~ N p,
n n
Y ± z ( α / 2 )⋅ p ⋅ (1 − p ) Y
= ± z ( α / 2 )⋅ σ
n n n
Y / n ⋅ (1 − Y / n ) Y ⋅( n − Y )
σ= =
n n3
e portanto:
214
Y−p
n ~ N ( 0,1 )
Y ⋅( n − Y )
n3
Y Y ⋅( n − Y ) Y + z ( α / 2 ) ⋅ Y ⋅( n − Y )
− z ( α / 2 ) ⋅ 3
, 3
n n n n
S2
( n − 1 )⋅ 2 ~ χ2n −1
σ
[( )
P χ 2n −1
A
(
< χ 2n −1 < χ 2n −1 ) ]= 1 − α
B
215
(
2
) S2
(
P χ n −1 < ( n − 1 )⋅ 2 < χ n −1 = 1 − α
A σ
2
B
)
ou:
1 σ 2
1
> > =1− α
( ) (χ )
P
χ n −1 A
2
( n − 1 )⋅ S2 2
n −1 B
( n − 1 )⋅ S ( n − 1 )⋅ S = 1 − α
2 2
> σ2 >
( ) ( )
P
χ n −1 A χ 2n −1
2
B
ou finalmente:
( n − 1 )⋅ S ( n − 1 )⋅ S = 1 − α
2 2
< σ2 <
( ) ( )
P
χ n −1 B χ n −1
2 2
A
216
( n − 1 )⋅ S ( n − 1 )⋅ S
2 2
( ) ( )
,
χ n −1 B χ n −1
2 2
A
(χ )
2
n −1 B = χ 2n −1 ( α / 2 )
(χ )
2
n −1 A = χ 2n −1 ( 1 − α / 2 )
2
( n − 1 )⋅ S , ( n − 1 )⋅ S
2
2
χ
n −1 ( α / 2 ) χ 2
n −1 ( 1 − α / 2 )
217
S2A
( nA − 1 ) ⋅ ~ χ 2n A −1
σ 2A
e
S2B
( nB − 1 ) ⋅ ~ χ 2n B −1
σB
2
resultando que:
S2A / σ 2A χ 2n A −1 / ( n A −1 )
~
S2B / σ 2B χ 2n B −1 / ( n B −1 )
S2A / σ 2A
~ Fn A −1 , n B −1
S2B / σ 2B
[ ]
P Fn A −1 , n B −1 ( 1− α / 2 ) < Fn A −1 , n B −1 < Fn A −1, n B −1 (α / 2 ) = 1 − α
e portanto:
S2A / σ 2A
P Fn A −1 , n B −1 ( 1 − α / 2 ) < 2 2 < Fn A −1 , n B −1 (α / 2 ) = 1 − α
SB / σ B
ou ainda:
1 σ 2A / σ 2B 1
P > 2 2 > =1− α
Fn A −1, n B −1 (1 − α / 2 ) SA / SB Fn A −1, n B −1 (α / 2 )
ou de outro modo:
1 S2A σ 2A 1 S2A
P ⋅ 2 > 2 > ⋅ 2 =1− α
Fn A −1, n B −1 (1 − α / 2 ) SB σ B Fn A −1, n B −1 (α / 2 ) SB
e finalmente:
1 S2A σ 2A 1 S2A
P ⋅ 2 < 2 < ⋅ 2 =1− α
Fn A −1, n B −1 (α / 2 ) SB σ B Fn A −1, n B −1 (1 − α / 2 ) SB
219
1 S2A 1 S2A
⋅ 2 , ⋅ 2
F
n A −1, n B −1 (α / 2 ) SB Fn A −1, n B −1 (1 − α / 2 ) SB
S2A ≈ σ 2A e S2B ≈ σ 2B
220
σ 2 S 2
XA ~ N µA , A ≈ N µA , A
n A n A
e
σ 2 S 2
XB ~ N µB , B ≈ N µB , B
n B n B
σ 2
σ 2 S 2
S 2
XA − XB ~ N µA − µB , A + B ≈ N µA − µB , A + B
nA nB nA nB
isto é:
Z=
( XA − XB ) − ( µA − µB ) ~ N ( 0,1 )
S2A S2B
+
nA nB
σ 2A = σ B
2
= σ2
então:
2 2
X A − X B ~ N µ A − µ B , σ + σ = N µ A − µ B , σ 2 1 + 1
nA nB nA nB
S = 2 (n A − 1)⋅S2A + (n B − 1)⋅S2B
nA + nB − 2
( X A − X B ) − z ( α / 2 ) ⋅ S ⋅
1 + 1 ,( X − X ) + z( α / 2 )⋅S⋅ 1 + 1
A B
nA nB nA nB
222
S2A ≈ σ 2A e S2B ≈ σ 2B
( XA − XB ) − ( µA − µB )
S2A S2B
+
nA nB
( XA − XB ) − ( µA − µB )
( XA − XB ) − ( µA − µB ) = σ 2A / n A + σ 2B / n B
S2A S2B S2A / n A + S2B / n B
+
nA nB σ 2A / n A + σ 2B / n B
N (0,1)
= ~ t gl
χgl
2
/ gl
gl = n A + n B − 2 se σ 2A = σ B
2
= σ2
2
S2A S2B
+
nA nB
gl = σ 2A ≠ σ 2B
(S ) + (S )
2 2
se
2 2
A / n A B / n B
nA − 1 nB − 1
S =
2 (n A − 1)⋅S2A + (n B − 1)⋅S2B
nA + nB − 2
224
σ 2A = σ B
2
= σ2
( µ A − µ B ) ∈ ( X A − X B ) ± t ( α / 2 )⋅S ⋅ 1 + 1
nA nB
σ 2A ≠ σ 2B
S2A S2B
( µA − µB ) ∈ ( X A − X B ) ± t ( α / 2 )⋅ +
nA nB
YA
amostras, seja um estimador de p A baseado numa
nA
Y
amostra de dimensão n A e B o estimador de p B baseado
nB
numa amostra de dimensão n B .
YA YB p ⋅ (1 − pA ) pB ⋅ (1 − pB )
− ~ N µ A − µ B , A +
nA nB nA nB
( pA − pB ) ∈
Y Y ⋅ ( n A − YA ) YB ⋅ ( n B − YB )
± z (α / 2 )⋅ A
Y
∈ A − B 3
+
nA nB nA n 3B
226
DIMENSIONAMENTO DE AMOSTRAS