Prague University of Economics and Business

FACULTY OF INFORMATICS AND STATISTICS

Department of Statistics and Probability

STATISTICS
FORMULAS FOR 4ST601

Version 3.54
Last Update: January 2024

KSTP, VŠE 2024

Descriptive statistics
nj k k
pj 
n
n j 1
j n p
j 1
j 1 j 1, 2, ..., k

x(1)  x(2)  ...  x( n )

xP , 0  P  1 xP  x( h )  (h   h  )( x( h )  x( h ) )
h  (n  1)  P  1

i 1, 2, ..., n
n k

 xi x n j j k
x   xj pj
i 1 j 1
x x k

n
n j 1
j
j 1

x n j j k
x   xj pj
j 1
x k

n
j 1
j
j 1

r  xmax  xmin rq  x0.75  x0.25

2
n n
 n 
 (x  x ) i
2
 xi2   xi 
 
i 1 i 1
s 
2 i 1
sx2  x 2  x 2 
 n 
x
n n
 
 
2
k
 k k

 (x j  x) nj 2
k  x n j   xj nj
2
j 
  (x j  x ) 2 p j  k 
j 1 j 1 j 1
s 
2
sx  x  x  k
2 2 2

 
x k

n j
j 1
 nj   nj 
j 1 j 1  j 1 
k k
  x 2j p j  ( x j p j ) 2
j 1 j 1
k k

s 2
jx nj  (x j  x )2 n j k k
  s 2jx p j   (x j  x ) 2 p j
j 1 j 1
sx2  s 2jx  sx2j  k
 k

n j 1
j n
j 1
j
j 1 j 1

sx  s 2
x

n n

 ( xi  x )2 n 2  (x  x ) i
2

s2  i 1
 sx s  s2  i 1

n 1 n 1 n 1

sx
mad  median( xi  x ) vx 
x

1
Descriptive analysis of dependence
Contingency table (r x c)
c r nij c
ni. r n. j
ni.   nij n. j   nij pij  pi.   pij  p. j   pij 
j 1 i 1 n j 1 n i 1 n
nij pij nij pij
pi| j   p j|i  
n. j p. j ni. pi.
ni.n. j r (nij  nij ) 2
c
nij  g  
n i 1 j 1 nij

g g
Cval   0, (m  1) / m Vval   0,1 m = min (r,c)
gn n(m  1)

One-way analysis of variance

nj nj

ssT    x ji  x  = ssTR + ssE ssTR    x j  x  n j ssE    x ji  x j 

k k k
2 2 2

j 1 i 1 j 1 j 1 i 1

ssTR
r2 
ssT

Correlation analysis
xy  x  y sxy
ryx  rxy  r  
x 2
 x2  y 2
 y2  sx s y

Regression analysis
Regresssion line ŷ  b0  b1 x yˆi  b0  b1 xi , i  1, 2,..., n
xy  x  y sxy
b1  byx   2 b0  y  b1x
x2  x 2 sx
Quadratic regression function ŷ  b0  b1 x  b2 x 2 yˆi  b0  b1 xi  b2 xi2 , i  1, 2,..., n
Regression plane ŷ  b0  b1 x1  b2 x2 yˆi  b0  b1 x1,i  b2 x2,i , i  1, 2,..., n
Residuals ei  yi  yˆi yi  yˆi  ei , i  1, 2,..., n

n n n n
ssT   (yi  y ) 2 ssR    yˆi  y  ssE   ei2   ( yi  yˆi ) 2
2

i 1 i 1 i 1 i 1

ssT  ssR  ssE

ssR n 1
r2  2
rADJ  1  (1  r 2 )
ssT n p

2
Descriptive time series analysis and predictions
n n 1
y1  y2 y  y3 y  yn
y
1 1
t y1   yt  yn d1  2 d 2  ...  n 1 d n 1
2 2
y
t 1
y
t 2
y 2 2 2
n n 1 d1  d 2  ...  d n 1
1 n y  y1
t = yt - yt-1  
n  1 t 2
t  n
n 1
yt yn
kt  k  n 1 k 2 k 3 ...k n  n 1
y t 1 y1
y yt I
I t /1  t  I 2 /1 I 3/ 2 ... I t / t 1 I t / t 1   t /1
y1 yt 1 I t 1/1

Time series decomposition

yt  trt  st  ct  et yt  trt  st  ct  et
Trend functions
n
Constant trend trt  b0 , t  1, 2,..., n b0   yt / n
t 1

Linear trend trt  b0  b1t , t  1, 2,..., n

Quadratic trend trt  b0  b1t  b2t 2 , t  1, 2,..., n
Predictions yˆt  trt , t  n  1, n  2,...

Moving averages
yt  p  ...  yt 1  yt  yt 1  ...  yt  p
m  2 p 1 yt  trt 
m
yt  p  ...  yt 1  yt  yt 1  ...  yt  p 1 yt  p 1  ...  yt 1  yt  yt 1  ...  yt  p

m  2p yt  trt  m m
2
Simple approach to seasonal component
nj s

 (sij  eij )  sd j

st  et  yt  trt sd j  i 1
sd j  sd j 
j 1

nj s
nj

 (s ij  eij )
s
st  et  yt / trt si j  i 1
si j  si j  s

 si
nj
j
j 1

Regression approach to seasonal component (linear trend, length of seasonality 4)

yt  trt  st  et  b0  b1t  sd1x1  sd 2 x2  sd3 x3  sd 4 x4  et b0  b0*  a

yt  trt  st  et  b0*  b1t  a1x1  a2 x2  a3 x3  et

a1  a2  a3
a sd j  a j  a , j  1, 2,3 sd 4   a
4
Predictions yˆt  trt  st  b0  b1  t  sd j , t  n  1, n  2,..., j  1, 2,3, 4

3
Probability and random variables
Probability theory
m
P(A)  P (A)  1  P (A)
n
P(A  B)  P(A)  P(B)  P(A  B)
Disjoint events P(A  B)  P(A)  P(B)
Independent events P(A)  P(A| B) P(B)  P(B | A) P(A  B)  P(A) P(B)
Dependent events P( A  B)  P( A)  P( B | A)  P( B)  P( A | B)

Random variables (random variable X; the value of this random variable x)

Discrete random variables
P(x)  P(X  x)  P(x)  1
x
F (x)  P(X  x)   P (t )
tx

P (x1  X  x2 )  F (x2 )  F (x1 )  

x1  x  x2
P (x )

E (X )   x P(x)
x

2
 
Var (X )  E  X 2    E  X     x 2 P (x)    x P (x) 
2
S.D.(X )  Var (X )
x  x 
Continuous random variables
 x
f (x)  F (x)  f ( x)dx  1

F (x)  P(X  x)  

f (t )dt

x2

P  x1  X  x2   F (x2 )  F (x1 )   f (x)dx

x1

xP , 0  P  1 F (xP )  P

E (X )   xf (x)dx


2

 
Var (X )  E  X    E  X     x f (x)dx    xf (x)dx 
2
2 2
S.D.(X )  Var (X )
   

4
The most important probability distributions
Probability distributions of discrete random variables
Binomial distribution: X ~ Bi(n,)
n
P( x )     x (1   ) n  x x = 0, 1, 2, ..., n, 0    1
 x
E ( X )  n Var ( X )  n 1   

Poisson distribution: X ~ Po()

x
P(x)  e   x = 0, 1, ... , >0
x!
E( X )   Var ( X )  

Hypergeometric distribution: X ~ Hy(N,M,n)

M  N  M 
   
 x   nx 
P( x)  , x  max(0, M -N +n),..., min(M , n) , n > 0, N ≥ n, M ≤ N
N
 
 n
M M  M  N n
E( X )  n Var (X )  n 1  
N N N  N 1

Probability distributions of continuous random variables

Standard normal distribution: Z ~ N(0,1), X ~ N(,2)
X  x
Z z   z  
 
E (Z )  0 Var ( Z )  1
 (z )  1   (  z ) z P   z1 P

Normal distribution: X ~ N(,2) -  x   -  µ   2 > 0

E( X )   Var ( X )   2

 x 
F ( x)   ( z )     xP     z P
  
 x   X   x2   
P (x1  X  x2 )  P  1    P(z1  Z  z2 )   (z2 )   (z1 )
    

Chi-square distribution: G ~ 2(df) g0

t-distribution (Student’s distribution): T ~ t(df)   t   t P (df )  t1 P (df )

1
F-distribution (Fisher – Snedecor d.): F ~ F(df1, df2) f 0 f P (df1 ,df 2 ) 
f1 P (df 2 ,df1 )

5
Statistical inference: point and interval parameter estimates

Calculated point population parameter estimates

Population mean (expected value)
Normal distribution in population; X ~ N(,2): ˆ = x
General distribution in population: Eˆ(X )  x

Population variance
Normal distribution in population; X ~ N(,2): ˆ 2  s 2
nj
Population proportion ˆ = p 
n
Count in population N   Np

Constructed (empirical) interval population parameter estimates

(confidence level 1 - )

Population mean (expected value)

Normal distribution in population, population variance 2 unknown, two-sided confidence
interval for parameter μ:
 s s 
 x  t1 / 2 ,x  t1 / 2  T ~ t(n – 1)
 n n

General distribution in population, Var(X) unknown (but finite), random sample size assumed
to be large enough, two-sided confidence interval for parameter E(X):
 s s 
 x  z1 / 2 ,x  z1 / 2 
 n n

Population proportion
Random sample size assumed to be large enough, two-sided confidence int. for parameter :
 p(1  p) p(1  p) 
 p  z1 / 2 , p  z1 / 2 
 n n 

Count in population
Random sample size assumed to be large enough, N known, two-sided confidence interval for
parameter N:
 p(1  p) p(1  p) 
N   p  z1 / 2 , p  z1 / 2 
 n n 

6
Statistical inference: statistical hypothesis tests

Statistical hypothesis tests of population parameters

Population mean (expected value); normal distribution in population, 2 unknown
H0 H1 Test statistic Value of test statistic Rejection region
 = 0  > 0 X  0 x  0 Wα={t; t  t1–}
T  n t  n
 < 0 S s Wα={t; t  –t1–}
  0 Wα={t; t  t1–/2}
Assuming H0 is true:
T ~ t(n – 1)

Population mean (expected value); general distribution in population, Var(X) unknown

(but finite), random sample size assumed to be large enough
H0 H1 Test statistic Value of test statistic Rejection region
E(X)= 0 X) > 0 X  0 x  0 Wα={z; z  z1–}
Z  n z  n
X < 0 S s Wα={z; z  –z1–}
X  0 Wα={z; z  z1–/2}
Assuming H0 is true:
Z ≈ N(0,1)

Population proportionrandom sample size assumed to be large enough

H0 H1 Test statistic Value of test statistic Rejection region
 = 0  > 0 P 0 p 0 Wα={z; z  z1–}
Z z
 < 0  0 (1   0 )  0 (1   0 ) Wα={z; z  –z1–}
  0 n n Wα={z; z  z1–/2}
Assuming H0 is true:
Z ≈ N(0,1)

Statistical hypothesis test of distribution in population

Chi-square goodness-of-fit test (random sample size and 0,j assumed to be large enough
nj  n 0, j

H0 and H1 Test statistic Value of test statistic Rejection region

H0: j = 0,j G k (n j  nj ) 2 Wα={g; g  21-}
(j = 1, .., k) Assuming H0 is true: g 
j 1 nj
H1: non H0 G ≈ 2(k – 1)

7
Inferential analysis of dependence, inferential time series analysis

Contingency table (r x c) (all expected frequencies assumed to be large enough

H0 H1 Test statistic Value of test statistic Rejection region
Variables are non H0 G r c (nij  nij ) 2 Wα ={g; g ≥ 21-}
g  
independent
Assuming H0 is true:
i 1 j 1 nij
G ≈ 2((r - 1)(c - 1))

One-way analysis of variance

H0 and H1 Test statistic Value of test statistic Rejection region
H0: 1 = 2 = … =k F ssTR
Wα={f; f  f1-α}
H1: non H0 Assuming H0 is true: f  k 1
ssE
F ~ F(k – 1, n – k) nk

Correlation analysis
H0 H1 Test statistic Value of test statistic Rejection region
ρXY = 0 ρXY  0 T r n2 Wα={t; t  t1–/2}
t
Assuming H0 is true: 1 r 2

T ~ t(n –2)

Regression line (p = 2), data-generating process:

Yi   0  1  xi   i i  1, 2,..., n
Hypothesis test of slope (t-test of slope), s(b1) only from Excel
H0 H1 Test statistic Value of test statistic Rejection region
1 = 0 1  0 T b Wα={t; t  t1–/2}
t 1
Assuming H0 is true: s (b1 )
T ~ t(n –p)

Linear trend (p = 2), data-generating process:

Yt   0  1  t   t t  1, 2,..., n
Hypothesis test of slope (t-test of slope), s(b1) only from Excel
H0 H1 Test statistic Value of test statistic Rejection region
1 = 0 1  0 T b Wα={t; t  t1–/2}
t 1
Assuming H0 is true: s (b1 )
T ~ t(n –p)

8
Index number analysis
c
All totals: i1

Q = pq IQ  Ip  Iq

p1 q1 Q1
Ip   p  p1  p0 Iq  q  q1  q0 IQ  Q  Q1  Q0
p0 q0 Q0

 q,  Q, p   q
Q
Homogeneous commodities


I (Σq) 
 q   Iq  q
1 0

q 1
Δ(q)   q1   q0
q q 0 0
q
 Iq 1

I (ΣQ) 
Q   p q 1 1 1

 IQ  Q   Q 0 1
(Q)   Q1   Q0
Q  p q 0 0 0 Q Q
 IQ
0 1

Q
 Q1  p1q1
1
Q
p 1

 q1   q1   p  p1  p0   pq
p p1q1
Ip  1  1
 0 0

p0  Q0  p0 q0 Q 0 q 1 q 0

 q0  q0 Q
p 0

Heterogeneous commodities

Ip ( L ) 
 p1q0

 Ip  p q 0 0

 Ip  Q 0
Ip ( P ) 
pq  pq
1 1 1 1

Q 1

 p0 q0  pq 0 0 Q 0 p q  pq
0 1 1 1 Q
 Ip 1
Ip

Ip ( F )  Ip ( L ) Ip ( P )

Iq ( L ) 
 p0 q1

 Iq  p q 0 0

 Iq  Q 0
Iq ( P ) 
 p1q1

pq 1 1

Q 1

 p0 q0 pq 0 0 Q 0  p1q0 pq
 Iq 1 1 Q
 Iq 1

Iq ( F )  Iq ( L ) Iq ( P )

9
Selected Excel functions
What Excel function Lecture Symbol
Frequency of a qualitative variables’ level COUNTIF() 1 nj
Frequencies of quantitative variables’ values FREQUENCY() 1 nj
Sum (total) SUM() 1 
Quantile PERCENTILE.INC() 1 xP
Median MEDIAN() 1 x
Mode MODE.MULT() 2 x̂
Simple arithmetic mean AVERAGE() 2 x
Variance (from raw data set) VAR.P() 2 sx2
Standard deviation (from raw data set) STDEV.P() 2 sx
Sample variance (from raw data set) VAR.S() 2 s2
Sample standard variation (from raw data set) STDEV.S() 2 s
Pearson’s correlation coefficient CORREL() 4 r
Probability function for X with binomial distribution BINOM.DIST() 8 P(x)
Probability function for X with hypergeometric distribution HYPGEOM.DIST() 8 P(x)
Probability function for X with Poisson distribution POISSON.DIST() 8 P(x)
Cumulative distribution function for Z NORM.S.DIST() 8  (z )
Quantiles for Z NORM.S.INV() 8 zP
Cumulative distribution function for X with normal
NORM.DIST() 8 F(x)
distribution
Quantiles for X with normal distribution NORM.INV() 8 xP
Quantiles for T with t-distribution T.INV() 8 tP
Quantiles for G with chi-squared distribution CHISQ.INV() 8  P2
Quantiles for F with F-distribution F.INV() 8 fP
p-value for the test of independence in the contingency table CHISQ.TEST() 11 p-value

10