Formula Sheet

Preliminaries:

2
∑𝑛𝑖=1(𝑋𝑖 − 𝑥̅ )2 𝑛 ∑𝑛𝑖=1 𝑋𝑖 2 − (∑𝑛𝑖=1 𝑋𝑖 )
𝑠= √ = √
𝑛−1 𝑛(𝑛 − 1)

Estimation: Single Population

Common Point Estimators under Simple Random Sampling Without Replacement

Parameter Point Estimator
Mean: Sample Mean:
𝜇 ∑𝑛𝑖=1 𝑋𝑖
𝑥̅ =
𝑛
Standard Error of the sample mean: Estimator of Standard Error of the sample mean:
𝑁 − 𝑛 𝜎2 𝑁 − 𝑛 𝑠2
𝑆𝐸(𝑋̅) = √( ) ̂ (𝑋̅) = √(
𝑆𝐸 )
𝑁−1 𝑛 𝑁 𝑛

Proportion: Sample Proportion:

p 𝑝̂
Standard Error of the sample proportion: Estimator of Standard Error of the sample
proportion:
𝑁 − 𝑛 𝑝(1 − 𝑝)
𝑆𝐸(𝑝̂ ) = √( ) 𝑁 − 𝑛 𝑝̂ (1 − 𝑝̂ )
𝑁−1 𝑛 ̂ (𝑝̂ ) = √(
𝑆𝐸 )
𝑁 𝑛−1

Common Point Estimators under Random Sampling from an Infinite Population

Parameter Point Estimator
Mean: Sample Mean:
𝜇 ∑𝑛𝑖=1 𝑋𝑖
𝑥̅ =
𝑛
Standard Error of the sample mean: Estimator of Standard Error of the sample mean:
𝜎2 𝑠2
𝑆𝐸(𝑋̅) = √ ̂ (𝑋̅) = √
𝑆𝐸
𝑛 𝑛

Proportion: Sample Proportion:

p 𝑝̂
Standard Error of the sample proportion: Estimator of Standard Error of the sample proportion:
𝑝(1 − 𝑝) 𝑝̂ (1 − 𝑝̂ )
𝑆𝐸(𝑝̂ ) = √ ̂ (𝑝̂ ) = √
𝑆𝐸
𝑛 𝑛
Confidence Intervals (Single Population)

1    100% Confidence Interval Estimates for the Population Mean

Cases Confidence Intervals

𝜎 𝜎
𝜎 2 is known (𝑥̅ − (𝑧𝛼 ) ( ) , 𝑥̅ + (𝑧𝛼 ) ( ))
2 √𝑛 2 √𝑛

𝑠 𝑠
𝜎 2 is unknown (𝑥̅ − (𝑡 𝛼, 𝑛−1
)( ) , 𝑥̅ + (𝑡 𝛼, 𝑛−1
)( ))
2 √𝑛 2 √𝑛

𝜎 2 is unknown but the sample size is 𝑠 𝑠

sufficiently large (𝑥̅ − (𝑧𝛼 ) ( ) , 𝑥̅ + (𝑧𝛼 ) ( ))
(Assume sufficiently large n if n>30) 2 √𝑛 2 √𝑛

An approximate 1    100% confidence interval estimate for the population proportion p when p is not
expected to be close to 0 or 1 and n is large:

𝑝̂(1−𝑝̂) 𝑝̂(1−𝑝̂)
(𝑝̂ − (𝑧𝛼 ) √ , 𝑝̂ + (𝑧𝛼 ) √ )
2 𝑛 2 𝑛
Confidence Intervals (Two Populations)

1    100% Confidence Interval Estimates for the Difference Between Two Population Means
(Independent Samples)
Cases Confidence Interval Estimates
σy 2 2 σy 2 2
𝜎𝑥 2 and σ
((𝑋̅ − 𝑌̅) − 𝑧𝛼 √ x +
σ
, (𝑋̅ − 𝑌̅) + 𝑧𝛼 √ x + )
𝜎𝑦 2 are 2 𝑛1 𝑛2 2 𝑛1 𝑛2
known

𝜎𝑥 2 and ̅) − (𝑡 𝛼 ) √sp 2 ( 1 + 1 ) , (X
̅−Y ̅) + (𝑡 𝛼 ) √sp 2 ( 1 + 1 ))
̅−Y
((X
𝜎𝑦 2 are , 𝑣
2 n n 1
, 𝑣 2 n n 2 1 2
unknown
(𝑛1 −1)𝑠𝑥 2 +(𝑛2 −1)𝑠𝑦 2
Where 𝑣 = n1 + n2 − 2 and sp 2 =
𝜎𝑥 2 = 𝜎𝑦 2 𝑛1 +𝑛2 −2

2 2 2 2
̅) − (𝑡 𝛼 ) √(𝑠𝑥 + 𝑠𝑦 ) , (X
̅−Y
((X ̅) + (𝑡 𝛼 ) √(𝑠𝑥 + 𝑠𝑦 ))
̅−Y
, 𝑣 n n 1
, 𝑣 n2 n 1 2
2 2
𝜎𝑥 2 and
𝜎𝑦 2 are 2 2
𝑠𝑥 2 𝑠𝑦
unknown ( + )
n1 n2
Where 𝑣= 2 2
𝜎𝑥 2 ≠ 𝜎𝑦 2 𝑠𝑥 2 𝑠𝑦 2
(n ) (n )
1 2
+
n1 −1 n2 −1

𝜎𝑥 2 and
𝜎𝑦 2 are
unknown 2 2 2 2
𝑠 𝑠𝑦 𝑠 𝑠𝑦
but ̅−̅
((X ̅−̅
Y) − 𝑧𝛼 √( 𝑥 + ) , (X Y) + 𝑧𝛼 √( 𝑥 + ))
n n2 1 n n
2 2 1 2
𝑛1 > 30
and
𝑛2 > 30

1    100% confidence interval estimate for the mean difference in matched or paired samples
𝑠𝑑 𝑠𝑑
(𝑑̅ − (𝑡𝛼 , 𝑛−1 ) √𝑛, 𝑑̅ + (𝑡𝛼 , 𝑛−1 ) √𝑛)
2 2

An approximate 1    100% confidence interval for p1  p 2 when the sample sizes are large

̂(1−𝑝
𝑝1 ̂)
1 ̂(1−𝑝
𝑝2 ̂)
2 ̂(1−𝑝
𝑝1 ̂)
1 ̂(1−𝑝
𝑝2 ̂)
2
((𝑝̂1 − 𝑝̂2 ) − 𝑧𝛼 √ + , (𝑝̂1 − 𝑝̂2 ) + 𝑧𝛼 √ + )
2 𝑛1 𝑛2 2 𝑛1 𝑛2
Tests of Hypothesis (Single Population)

Summary of Hypothesis Tests on the Population Mean

Alternative
Null Hypothesis Region of
Hypothesis Test Statistic
Ho Rejection
Ha
 is known
  0 z   z
X  o
  0   0 Z z  z

  0 n z  z 2
 is not known
  0 𝑡 < −𝑡 𝛼, 𝑛−1
X  o
  0   0 t 𝑡 > 𝑡 𝛼, 𝑛−1
S
  0 n |𝑡| > 𝑡 𝛼, 𝑛−1
2
 is not known but the sample size is sufficiently large
(Assume sufficiently large n if n>30)
  0 𝑡 < −𝑧𝛼
X  o
  0   0 t 𝑡 > 𝑧𝛼
S
  0 n |𝑡| > 𝑧𝛼
2

Hypothesis Test on the Population Proportion (For large n)

Alternative
Null Hypothesis Region of
Hypothesis Test Statistic
Ho Rejection
Ha
p  po X  np o z   z
Z
np o (1  p 0 )
p  po p  po z  z
where
p  po z  z 2
X = number of success
Tests of Hypothesis (Two Populations)

Summary of Tests of Hypotheses for the Difference of Means

Alternative
Null Region of
Hypothesis Test Statistic
Hypothesis Ho Rejection
Ha
 X2 and  Y are known
2

 X  Y  d o z   z
( X  Y )  do
Z
 X  Y  d o  X  Y  d o  X2  Y2 z  z

n1 n2
 X  Y  d o z  z 2

 X2   Y2   2 is unknown
( X  Y )  do
 X  Y  d o t 𝑡 < −𝑡 𝛼, 𝑣
1 1
Sp 
n1 n2
 X  Y  d o  X  Y  d o where 𝑡 > 𝑡 𝛼, 𝑣
(n1  1) S  (n2  1) S
2 2
S p2  X Y

n1  n2  2 |𝑡| > 𝑡𝛼, 𝑣

 X  Y  d o
v  n1  n2  2 2

 X2   Y2 and both are unknown

(X  Y )  do
 X  Y  d o t 𝑡 < −𝑡 𝛼, 𝑣
S X2 S Y2

n1 n 2
where
 X  Y  d o  X  Y  d o 2 𝑡 > 𝑡 𝛼, 𝑣
 S X2 S2 
  Y
 n1 n2 
v 2 2
 S X2   S Y2 
 X  Y  d o  
n1   n  |𝑡| > 𝑡𝛼, 𝑣
 
2 
2
n1  1 n2  1
 X2 and  Y2 are unknown but 𝑛1 > 30 and 𝑛2 > 30

 X  Y  d o z   z
( X  Y )  do
Z
 X  Y  d o  X  Y  d o S X2 S Y2 z  z

n1 n2
 X  Y  d o z  z 2
 Test for the difference of population means for paired or matched observations

Alternative
Null Hypothesis Region of
Hypothesis Test Statistic
Ho Rejection
Ha
D  do 𝑡 < −𝑡 𝛼, 𝑛−1
d  do
D  do t 𝑡 > 𝑡 𝛼, 𝑛−1
 D   X  Y  d o Sd
D  do n |𝑡| > 𝑡 𝛼, 𝑛−1
2

2
∑𝑛𝑖=1 𝑑𝑖 ∑𝑛𝑖=1(𝑑𝑖 − 𝑑̅ )2 𝑛 ∑𝑛𝑖=1 𝑑𝑖 2 − (∑𝑛𝑖=1 𝑑𝑖 )
𝑑̅ = , 𝑆𝑑 = √ = √
𝑛 𝑛−1 𝑛(𝑛 − 1)

Hypothesis Test on the Difference of Two Population Proportions ( for 𝑙𝑎𝑟𝑔𝑒 𝑛1 𝑎𝑛𝑑 𝑛2 )

Null Alternative
Region of
Hypothesis Hypothesis Test Statistic
Rejection
Ho Ha
p1  p 2 𝑝
̂1 − 𝑝
̂2 z   z
𝑍=
1 1
√𝑝̅ (1 − 𝑝̅ ) (
𝑛1 + 𝑛2 )
p1  p 2 z  z
p1  p 2 where
𝑥1 + 𝑥2
𝑝̅ =
𝑛1 + 𝑛2
p1  p 2 z  z 2
𝑥𝑖 = 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑠𝑢𝑐𝑐𝑒𝑠𝑠
𝑛𝑖 = 𝑠𝑎𝑚𝑝𝑙𝑒 𝑠𝑖𝑧𝑒

Test for Independence

Ho: The variables are independent

Ha: The variables are not independent

Test statistic:
𝑟 𝑐 2 2 𝑟 𝑐
2
(𝑂𝑖𝑗 − 𝐸𝑖𝑗 ) 𝑂𝑖𝑗
𝑋 = ∑∑ = ∑∑ −𝑛
𝐸𝑖𝑗 𝐸𝑖𝑗
𝑖=1 𝑗=1 𝑖=1 𝑗=1
th
where Oij is the observed number of cases in the i row and jth column
Eij is the expected number of cases in the ith row and jth column

Region of rejection: 𝑋 2 > 𝜒 2 𝛼, (𝑟−1)∗(𝑐−1)

 Correlation Coefficient

n
 n  n 
n X i Yi    X i   Yi 
r i 1  i 1  i 1 
 n 2  n 
2
 n 2  n  2 
 n X    X   n Y    Y  
 i 1 i  i 1 i   i 1 i  i 1 i  
  

Alternative
Null Hypothesis Region of
Hypothesis Test Statistic
Ho Rejection
Ha
  0 𝑡 < −𝑡𝛼, 𝑛−2
  0 r   0  n2 𝑡 > 𝑡𝛼, 𝑛−2
  0 t
1 r2
  0 |𝑡| > 𝑡𝛼, 𝑛−2
2

Regression (Least Square Estimates)

n
 n  n 
n X i Yi    X i   Yi 
b1  i 1  i 1  i 1 
2 ,
n
 n 
n X i    X i 
2

i 1  i 1 

bo  Y  b1 X .
̂𝑖 = 𝑏0 + 𝑏1 𝑋𝑖
𝑌
2
2
𝑆𝑆𝐸 ∑𝑛𝑖=1(𝑌𝑖 − 𝑌̂𝑖 )
𝜎̂ = 𝑀𝑆𝐸 = =
𝑛−2 𝑛−2
 A (1 − 𝛼)100% confidence interval for 𝛽0 is given by:

(𝑏0 − (𝑡𝛼 , 𝑛−2 )𝑆𝑏0 , 𝑏0 + (𝑡𝛼, 𝑛−2 )𝑆𝑏0 )

2 2

(𝑀𝑆𝐸)(∑𝑛 2
𝑖=1 𝑥𝑖 )
Where 𝑆𝑏0 =√ 2
𝑛 ∑𝑛 2 𝑛
𝑖=1 𝑥𝑖 −(∑𝑖=1 𝑥𝑖 )

A (1 − 𝛼)100% confidence interval for 𝛽1 is given by:

(𝑏1 − (𝑡𝛼 , 𝑛−2 )𝑆𝑏1 , 𝑏1 + (𝑡𝛼 , 𝑛−2 )𝑆𝑏1 )

2 2

(𝑀𝑆𝐸)(𝑛)
Where 𝑆𝑏1 =√ 2
𝑛 ∑𝑛 2 𝑛
𝑖=1 𝑥𝑖 −(∑𝑖=1 𝑥𝑖 )

Test for 𝛽0

Ho: 𝛽0 = 0 vs. Ha: 𝛽0 ≠ 0

𝑏0
Test statistic: 𝑡 =
𝑆𝑏0

Reject the null hypothesis if: |𝑡| > 𝑡 𝛼, 𝑛−2

2

Test for 𝛽1

Ho: 𝛽1 = 0 vs. Ha: 𝛽1 ≠ 0

𝑏1
Test statistic: 𝑡 =
𝑆𝑏1

Reject the null hypothesis if: |𝑡| > 𝑡 𝛼, 𝑛−2

2