Descriptive statistics

Frequency distributions and related

statistics
 Types of datasets
Separately registered data – ungrouped data

Frequency distribution – grouped data by

classes
Ungrouped, raw Grouped data
data (frequency distribution)

Observations Classes Frequency of

1 1 x occurrence
f
2 3
1 6
1 1
2 3
2 4
3 3
5 1
4 1
3 3
5 1
2 1
• Discrete frequency distribution
– Classes are separate, usually whole numbers
• Continuous frequency distribution
– Classes are continuous variables, that are limited
to a certain range
– Observations, that fall within one range, are
considered as similar
 Example
Discrete
0 1 2 3 4
€9,99 €10,99 €19,99

Continuous
(0-100] (100-200] (200-300]
(300-500] (500-1000]
Continuous frequency distribution
• Class interval length

i = x − x ''
i
'
i
• Usually same-length classes are used
• Constructed based on socio-economic context
Continuous frequency distribution
• Different class interval length
– if a classification system is used,
– if there are few observations in outlining intervals,
– if empty intervals appear.
Continuous frequency distribution
• Interval density – proportion of interval
frequency and length
Determining the length of intervals
• Sturges’ formula (n ≤ 100)
xmax − xmin
=
1 + 3,2  lg n
• Brex formula (n > 100)

xmax − xmin
=
5  lg n
Cumulative frequency distribution
• Used for analysis of total frequency up to the
class.
• The frequency of observations corresponding
to the value in the class or a lower value.
 Charts
• Polygon
– Used for discrete frequency distributions
– Horizontal axis — classes, vertical axis — frequencies
• Histogram
– Used for continuous frequency distributions
– Horizontal axis — segments, that correspond to length
of classes, vertical axis — frequencies
• Cumulate
– Cumulative frequency distribution
– Horizontal axis — values of discrete or continuous
classes, vertical axis — accumulated frequencies
 Polygon
(Scatter with lines)
Histogram
Analysis of a discrete frequency
distribution
 Mean statistics

• Objective
• Abstract
• Describe the phenomenon as a
whole
 Mean statistics
Measures of central tendency
Summary means
and
Structural (location) means

Simple or unweighed means –

for ungrouped data
and
Weighed means – for grouped data
 Arithmetic mean
• Simple arithmetic mean

x=
 x i

N
• Weighed arithmetic mean

x=
 x f i i

f i
 Arithmetic mean
𝒙𝒊
𝒙𝒊 𝒇𝒊
3
3 1
5
5 2
5
8 1
8

𝟑+𝟓+𝟓+𝟖 𝟑+𝟓∙𝟐+𝟖
ഥ
𝒙= ഥ
𝒙=
𝟒 𝟒
 Structural means
• Median
• Quantiles
• Mode
 Median
• Class, that separates an ordered frequency
distribution (ascending or descending) in two
equal parts (by frequencies).

• If the number of observations is a odd

number, the median is arithmetic mean from
two middle observations.
 Median
𝒙𝒊 8 7 3 9 1 2 2

𝒙𝒊 1 2 2 3 7 8 9

Me
 Mode
• The most frequent observation or class.
• x with the largest f
• For continuous frequency distributions:

f Mo − f Mo−1
Mo = x0 +  Mo
f Mo − f Mo−1 + f Mo − f Mo+1
 Mode
• One mode — monomodal
• Two modes — bimodal
• Three or more modes — multimodal
Variance statistics
How much observed values differ
from the average and how
significant are this differences.
 Variance statistics
• Variation range
• Variance and standard deviation
• Variance coefficient
 Variation range
• Different between the largest and smallest
observation

Rv = xmax − xmin
• Takes rare, extreme values into account!
 Variance and standard deviation
• Variance — mean quadratic deviation from the
arithmetic mean in quadratic measures
N

 (x − x )
2
i
 =
2 i =1

N
• Weighed variance

 (x − x ) 2
fi
 2
= i

f i
 Variance and standard deviation
𝒙𝒊 𝒙𝒊 − 𝒙ഥ 𝟐 𝒙𝒊 𝒇𝒊 𝒙 𝟐 𝒇𝒊
𝒙𝒊 − ഥ
3 (3 - 5,25)2
3 1 (3 - 5,25)2
5 (5 - 5,25)2
5 2 (5 - 5,25)2·2
5 (5 - 5,25)2
8 1 (8 - 5,25)2
8 (8 - 5,25)2
 Variance and standard deviation
• Standard deviation — mean quadratic deviation
from the arithmetic mean in the same measures
as xi

=  2
 Variance application
• Dispersion comparison
• Inequality analysis
• Convergence analysis
• What is “normal”?
• etc.
 Variance coefficient
• Relative level of variance (in percent)


V= 100
x
• Allows comparison of different objects with
different measures
Skewness and kurtosis statistics
 Skewness
• Skewness statistics characterise the skew of
the symmetry relatively to the arithmetic
mean
 Structural skewness
• An approximate statistic of skewness

x − Me
A=
x − Mo
• A = 0 → symmetric
• A < 3 → asymmetric
 Skewness coefficient
• A more precise skewness statistic

m3
K3 =
• K3 > 0 → positive skew
 3

• K3 < 0 → negative skew

 Kurtosis coefficient
• The concentration of observations around the
arithmetic mean, in comparison to the normal
distribution.

m4
E= −3
 4
• E > 0 → pointed at peak
• For normal distribution E = 0
 Central moments
n

 i
( x − x ) f i
k

mk = i =1
n


i =1
fi
Analysis of a continuous
frequency distribution
 Arithmetic mean
• Simple arithmetic mean

x=
 x i

N
• Weighed arithmetic mean

x=
 x f i i

f i
 Median
For continuous frequency distributions:
𝑛
σ𝑖=1 𝑓𝑖
σ 𝑀𝑒−1
− 𝑖=1 𝑓𝑖
2
𝑀𝑒 = 𝑥0 + Δ𝑀𝑒
𝑓𝑀𝑒
𝑥0 - start of the median interval
Δ𝑀𝑒 - length of the median interval
σ𝑛𝑖=1 𝑓𝑖 - total number of observations
σ𝑀𝑒−1
𝑖=1 𝑓𝑖 - number of observations until median
interval
𝑓𝑀𝑒 - number of observations in the median interval
 Mode
For continuous frequency distributions:
f Mo − f Mo−1
Mo = x0 +  Mo
f Mo − f Mo−1 + f Mo − f Mo+1
𝑥0 - start of the mode interval
Δ𝑀𝑜 - length of the median interval
𝑓𝑀𝑜 - number of observations in the mode interval
𝑓𝑀𝑜−1 - number of observations in the interval before the
mode interval
𝑓𝑀𝑜+1 - number of observations in the interval after the
mode interval
 Variance and standard deviation
• Variance — mean quadratic deviation from the
arithmetic mean in quadratic measures
N

 (x − x )
2
i
 =
2 i =1

N
• Weighed variance

 (x − x ) 2
fi
 2
= i

f i
Analysis of a growth rate
 Geometric mean
• Simple geometric mean
N
x0 = N  xi
i =1
• Weighed geometric mean
n

 n

 i
f
x0 =
i fi
i =1 x
i =1
 Example – growth rate vs. percent

× 100 𝑡ℎ𝑒𝑛 − 100

Growth rate Change in percent
1.34 +34%
+100 𝑡ℎ𝑒𝑛 ÷ 100

× 100 𝑡ℎ𝑒𝑛 − 100

Growth rate Change in percent
0.72 -28%
+100 𝑡ℎ𝑒𝑛 ÷ 100