Statistics for Management and Economics, Eighth Edition

Formulas

Numerical Descriptive techniques

Population mean

 = x i 1
i

Sample mean

x i 1
i

x
n

Range

Largest observation - Smallest observation

Population variance

2 = ( x   )
i 1
i
2

Sample variance

s2 =
i 1
 (x i  x)2

n 1
Population standard deviation

 = 2

Sample standard deviation

s= s2

Population covariance
 N

( x
i 1
i   x )( y i   y )
 xy 
N

Sample covariance

( x
i 1
i  x )( y i  y )
s xy 
n 1

Population coefficient of correlation

 xy

 x y

Sample coefficient of correlation

s xy
r
sx s y

Coefficient of determination

R2 = r2

Slope coefficient

s xy
b1 
s x2

y-intercept

b0  y  b1 x

Probability

Conditional probability

P(A|B) = P(A and B)/P(B)

Complement rule

P( A C ) = 1 – P(A)
Multiplication rule

P(A and B) = P(A|B)P(B)

Addition rule

P(A or B) = P(A) + P(B) - P(A and B)

Bayes’ Law Formula

P(A i )P(B | A i )
P(A i | B) 
P(A 1 )P(B | A 1 )  P(A 2 )P(B | A 2 )  . . .  P(A k )P( B | A k )

Random Variables and Discrete Probability Distributions

Expected value (mean)

E(X) =
  xP( x )
all x

Variance

V(x) =
2  ( x   )
all x
2
P( x )

Standard deviation

 2

Covariance

COV(X, Y) = σxy = ( x   x )( y   y )P( x , y )

Coefficient of Correlation

COV ( X , Y )

 x y

Laws of expected value

1. E(c) = c

2. E(X + c) = E(X) + c

3. E(cX) = cE(X)

Laws of variance

1.V(c) = 0

2. V(X + c) = V(X)

3. V(cX) = c 2 V(X)
Laws of expected value and variance of the sum of two variables

1. E(X + Y) = E(X) + E(Y)

2. V(X + Y) = V(X) + V(Y) + 2COV(X, Y)

Laws of expected value and variance for the sum of more than two variables

k k
1. E ( X i )   E ( X i )
i 1 i 1

k k
2. V ( X i )  V ( X ) if the variables are independent
i
i 1 i 1

Mean and variance of a portfolio of two stocks

E(Rp) = w1E(R1) + w2E(R2)

V(Rp) = w12 V(R1) + w22 V(R2) + 2 w1 w 2 COV(R1, R2)

= w12 12 + w22  22 + 2 w1 w 2  1  2

Mean and variance of a portfolio of k stocks

k
E(Rp) =  w E(R )
i 1
i i

k k k
V(Rp) = w
i 1
2 2
i i 2  w w
i 1 j i 1
i j COV ( Ri ,Rj)

Binomial probability

n!
P(X = x) = p x ( 1  p )n x
x! ( n  x )!

  np

 2  np( 1  p )

  np( 1  p )

Poisson probability

e   x
P(X = x) =
x!
Continuous Probability Distributions

Standard normal random variable

X 
Z


Exponential distribution

    1/ 

P( X  x )  e  x

P ( X  x )  1  e  x

P ( x 1  X  x 2 )  P ( X  x 2 )  P ( X  x 1 )  e   x1  e   x 2

F distribution

1
F1 A, 1 , 2 =
FA, 2 , 1

Sampling Distributions

Expected value of the sample mean

E( X )   x  

Variance of the sample mean

2
V ( X )   2x 
n

Standard error of the sample mean


x 
n

Standardizing the sample mean

X 
Z 
/ n

Expected value of the sample proportion

ˆ )   pˆ  p
E (P

Variance of the sample proportion

 p (1  p )
V ( Pˆ )   2pˆ 
n

Standard error of the sample proportion

p (1  p )
 pˆ 
n

Standardizing the sample proportion

ˆ p
P
Z 
p (1  p ) n

Expected value of the difference between two means

E ( X 1  X 2 )   x1  x2   1   2

Variance of the difference between two means

 12  22
V ( X 1  X 2 )   x21  x2  
n1 n2

Standard error of the difference between two means

12  22
 x1  x2  
n1 n 2

Standardizing the difference between two sample means

( X 1  X 2 )  ( 1   2 )
Z
 12  22

n1 n 2

Introduction to Estimation

Confidence interval estimator of 


x  z / 2
n

Sample size to estimate 

2
z  
n    / 2 
 W 
Introduction to Hypothesis Testing

Test statistic for 

x
z
/ n

Inference about One Population

Test statistic for 

x
t
s/ n

Confidence interval estimator of 

s
x  t / 2
n

Test statistic for  2

( n  1 )s 2
2 
2

Confidence interval Estimator of  2

( n  1 )s 2
LCL =
 2 / 2

( n  1 )s 2
UCL =
 12 / 2

Test statistic for p

p̂  p
z
p( 1  p ) / n

Confidence interval estimator of p

p̂  z / 2 p̂( 1  p̂ ) / n

Sample size to estimate p

 2
z p̂( 1  p̂ ) 
n /2
 W 
 

Confidence interval estimator of the total of a large finite population

 s 
N  x  t / 2 
 n

Confidence interval estimator of the total number of successes in a large finite population

 p̂( 1  p̂ 
N  p̂  z / 2 
 n 

Confidence interval estimator of  when the population is small

s N n
x  t / 2
n N 1

Confidence interval estimator of the total in a small population

 s N n 
N  x  t / 2 
 n N 1 
 

Confidence interval estimator of p when the population is small

 p̂( 1  p̂ ) N n
p  z / 2
n N 1

Confidence interval estimator of the total number of successes in a small population

 p̂( 1  p̂ ) N n 
N  p̂  z  / 2 
 n N 1 
 

Inference About Two Populations

Equal-variances t-test of 1   2

( x1  x 2 )  ( 1   2 )
t
 1 1    n1  n 2  2
s 2p   
 n1 n 2 

Equal-variances interval estimator of 1   2

  1 1 
( x1  x 2 )  t  / 2 s 2p      n1  n 2  2
 n1 n 2 

Unequal-variances t-test of 1   2

( x1  x 2 )  ( 1   2 )
t ( s12 / n1  s22 / n2 )2

 s12 s 22  ( s12 / n1 )2 ( s22 / n2 )2
   
 n1 n 2  n1  1 n2  1
 

Unequal-variances interval estimator of 1   2

( s12 / n1  s22 / n2 )2
s12 s 22 
( x1  x 2 )  t  / 2  ( s12 / n1 )2 ( s22 / n2 )2
n1 n 2 
n1  1 n2  1

t-Test of  D

xD   D
t   nD  1
s D / nD

t-Estimator of  D

sD
x D  t / 2   nD  1
nD

F-test of 12 /  22

s12
F=  1  n1  1 and  2  n 2  1
s22

F-Estimator of 12 /  22

 s12  1
LCL =  2 
s  F / 2 , ,
 2  1 2

 s12 
UCL =  2  F / 2 , ,
s  2 1
 2 

z-Test and estimator of p1  p 2

 ( p̂1  p̂ 2 )
z
Case 1:  1 1 
p̂( 1  p̂ )  
 n1 n 2 

( p̂1  p̂2 )  ( p1  p2 )
z
Case 2: p̂1( 1  p̂1 ) p̂2 ( 1  p̂2 )

n1 n2

z-estimator of p1  p 2

p̂1( 1  p̂1 ) p̂2 ( 1  p̂2 )

( p̂1  p̂2 )  z / 2 
n1 n2

Analysis of Variance

One-way analysis of variance

k
SST = n (x
j 1
j j  x )2

k nj

SSE =  ( x
j 1 i 1
ij  x j )2

SST
MST =
k 1

SSE
MSE =
nk

MST
F=
MSE

Two-way analysis of variance (randomized block design of experiment)

k b
SS(Total) =  ( x
j 1 i 1
ij  x )2

k
SST =  b( x [ T ]
i 1
j  x )2

b
SSB =  k( x [ B ]
i 1
i  x )2
 k b
SSE =  ( x
j 1 i 1
ij  x [ T ] j  x [ B ]i  x )2

SST
MST =
k 1

SSB
MSB =
b 1

SSE
MSE =
n  k  b 1

MST
F=
MSE

MSB
F=
MSE

Two-factor experiment

a b r
SS(Total) =   ( x
i 1 j 1 k 1
ijk  x )2

a
SS(A) = rb ( x [ A ]
i 1
i  x )2

b
SS(B) = ra ( x [ B ]
j 1
j  x )2

a b
SS(AB) = r ( x [ AB ]
i 1 j 1
ij  x [ A ]i  x [ B ] j  x )2

a b r
SSE =   ( x
i 1 j 1 k 1
ijk  x [ AB ]ij )2

MS( A )
F=
MSE

MS( B)
F=
MSE

MS(AB)
F=
MSE

Least Significant Difference Comparison Method

  1 1 
LSD = t  / 2 MSE  
 ni n j 
 

Tukey’s multiple comparison method

MSE
  q ( k , )
ng

Chi-Squared Tests

Test statistic for all procedures

k
(f i  e i ) 2
2  i 1
ei

Simple Linear Regression

Sample slope

s xy
b1 
s x2

Sample y-intercept

b0  y  b1 x

Sum of squares for error

n
SSE = ( y
i 1
i  ŷ i ) 2

Standard error of estimate

SSE
s 
n2

Test statistic for the slope

b1   1
t
s b1

Standard error of b1
 s
s b1 
( n  1 )s x2

Coefficient of determination

2
s xy SSE
R2   1
s x2 s 2y ( y i  y )2

Prediction interval

2
1 ( xg  x )
ŷ  t  / 2 ,n  2 s  1  
n ( n  1 )s x2

Confidence interval estimator of the expected value of y

2
1 ( xg  x )
ŷ  t  / 2 ,n  2 s  
n ( n  1 )s x2

Sample coefficient of correlation

s xy
r
sx s y

Test statistic for testing  = 0

n2
tr
1 r 2

Multiple Regression

Standard Error of Estimate

SSE
s 
n  k 1

Test statistic for  i

bi   i
t
s bi
Coefficient of Determination

2
s xy SSE
R  2  1
s x2 s 2y ( y i  y )2

Adjusted Coefficient of Determination

2 SSE /( n  k  1 )
Adjusted R  1 
 ( yi  y )2 /( n  1 )

Mean Square for Error

MSE = SSE/k

Mean Square for Regression

MSR = SSR/(n-k-1)

F-statistic

F = MSR/MSE

Durbin-Watson statistic

( e
i 2
i  ei 1 ) 2
d n

e
i 1
2
i

Nonparametric Statistical techniques

Wilcoxon rank sum test statistic

T  T1

n1 ( n1  n 2  1 )
E(T) =
2

n1 n 2 ( n1  n 2  1 )
T 
12

T  E( T )
z
T

Sign test statistic

 x = number of positive differences

x  .5 n
z
.5 n

Wilcoxon signed rank sum test statistic

T T

n( n  1)
E(T) =
4

n( n  1 )( 2n  1 )
T 
24

T  E( T )
z
T

Kruskal-Wallis Test

 12 k
T j2 
H 
 n( n  1 )
 nj 
  3( n  1 )
j 1 

Friedman Test

 12
k 
Fr  
 b( k )( k 1) T
j 1
j 
2

 3b( k  1 )

Spearman rank correlation coefficient

s ab
rS 
s a sb

Spearman Test statistic for n > 30

z  rS n 1

Time Series Analysis and Forecasting

Exponential smoothing

S t  wy t  (1  w) S t 1
Statistical Process Control

Centerline and control limits for x chart using S

Centerline = x

S
Lower control limit = x  3
n

S
Upper control limit = x  3
n

Centerline and control limits for the p chart

Centerline = p

p( 1  p )
Lower control limit = p  3
n

p( 1  p )
Upper control limit = p  3
n

Decision Analysis

Expected Value of perfect Information

EVPI = EPPI - EMV*

Expected Value of Sample Information

EVSI = EMV' - EMV*