You are on page 1of 8

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/275648121

Practical application of coefficient of variation

Article · October 2013

CITATIONS READS

0 838

1 author:

Marek Marcisz
Silesian University of Technology
58 PUBLICATIONS   53 CITATIONS   

SEE PROFILE

All content following this page was uploaded by Marek Marcisz on 30 April 2015.

The user has requested enhancement of the downloaded file.


Volver al índice

Practical application of coefficient of variation


Marcisz, Mareka
a
Silesian University of Technology, Faculty of Mining and Geology, Institute of Applied Geology, PL 44-100
Gliwice, 2 Akademicka Str., tel: +4832 2371026, fax: +4832 2372290, email: marek.marcisz@polsl.pl

ABSTRACT
The role and meaning of the coefficient of variation in mining geology and ore geology has been presented. The
study was conducted in conditions of bituminous coal deposits of the Upper Silesian Coal Basin (Poland). The
results were applied to moisture content Wa, ash content Ad, volatile matter content Vdaf, total sulphur content Std
and calorific value CVdaf. Based upon the obtained values of the coefficient of variation, these parameters were
assigned to adequate groups of variation. Density of sampling (distance between points of channel samples
collection) was calculated as well. The necessary number of samples for the estimation of resources in block of
deposit and number of control samples were also determined. The correction of the obtained results has been
performed. The results were adapted for conditions of underground mining of bituminous coal in the mines of
the Upper Silesian Coal Basin.

KEYWORDS: coefficient of variation, bituminous coal, quality parameters, density of sampling, Upper Silesian
Coal Basin

1. INTRODUCTION
The variation of mineral deposit is understood as the diversification values assumed by parameters in various
points of the deposit space - that is the properties of the deposit which may be measured in terms of quantity.
The source of the information concerning the deposit parameters is the data obtained by sampling of boreholes,
headings and the output material. The description and the appraisal of the deposit parameter values are provided
by mathematical statistic methods which utilize the calculus of probability. The method allows to draw
conclusions concerning the properties of the deposit, understood as a total population (a set of all the samples
which can be obtained from the bed), based upon a statistical sample (the sample population), understood as a set
of parameter measurement results assigned to samples. It is also assumed that the deposit parameters are random
variables (which, as a result of test, assume one of the values of a given set of real numbers with a certain
probability), characterized by distribution of random variable. It is assumed (in classical statistics), that the
analyzed variable does not show autocorrelation (Mucha J. 1994).
The preparation of the results of the deposit parameters measurement is aimed to replace the extensive data
obtained by sampling with a value of a statistical parameter (or parameters) which reflects the basic
characteristics.
The "V" coefficient of variation is one of these statistical parameters, a classical measure of the diversification of
characteristics distribution, which, in contrast to the standard deviation (determining the non-relative
diversification of a characteristic), is a relative measure, dependent upon the arithmetic mean value (formula 1):

s
s
V = *100%
(1)
where:
s - standard deviation
x
- arithmetic mean of the sample.
The coefficient of variation, which numerically characterizes the variation of the deposit parameters, may vary
for the deposit of one kind (even within the same deposit) in a very wide range, from several up to few hundred
percent. The variation of a deposit for a large area is usually higher that for its part. The coefficient of variation
values determined based upon coarse sampling grid are lower than these determined based upon fine sampling
grid. The coefficients of variation at small parts of the deposit (up to few ha), are often diversified to an extent
that areas of high and low variation may be distinguished (and presented on a map). This is usually a result of a
diversified geological structure of the deposit in such areas (Nieć M., 1982).
The coefficient of variation as a measure of the variation of deposit parameters may only be applied if these
parameters are measured using a ratio scale, in which the majority of the deposit parameters is measured, that is,
when no values lower than zero may occur. In situations where the parameter is measured in an interval scale
(such as floor elevation above sea level), the coefficient of variation may not be applied (the variation is then
determined by means of mean square deviation).
Volver al índice

It is one of the most commonly used summary statistic and it may also be defined as a relative measure of
dispersion of deposit parameter value, which provides the standard deviation expressed as a percentage of the
mean value of the parameter. Due to the above, it is possible to effectively compare parameters of various types,
such as strata thickness and the content of the useful component as well as to compare the variation of two
parameters of highly diversified mean values. The coefficient of variation plays a significant role in the design of
the borehole grid density, which should be adjusted to the deposit parameter of the highest variation (Mucha J.
1994).
This is due to the fact that the precision of identification "e" expressed as a percentage of the mean value is
dependent on the coefficient of variation (formula 2):

zV
ZV
e% =
n
(2)
where:
z - the probability coefficient dependent upon the assumed confidence level,
n - number of boreholes.

If the parameter has a normal distribution, formula 3 is applied:

V
*% =
" (3)
where :
t - Student's distribution (Gosset's distribution),
n - number of boreholes.

In the geology of deposits, the coefficient of variation constitutes a basis of statistical (quantitative) classification
of deposit variation, proposed by Baryszew (Table 1, columns 1-3).
The distinguished groups of deposit variation (Table 4, column 4) regulate the density of workings while
estimating the reserves of the deposits, preparing geological documentation and the scales of the maps of
reserves.
Representative sampling of a deposit, stratum or their parts, requires acquiring a correct number of samples,
based on which it will be possible to estimate the mean values of qualitative parameters of the mineral with the
required accuracy or the possible accuracy at a given variation of the deposit. The "l" recommended distance
between the subsequent sampling points may be determined using formula 4 (Nieć M. 1982):

l = 300 [m]
V
(4)

The coefficient of variation also plays a significant role in the estimation of the useful mineral in the deposit. It
allows to determine the number of controls necessary to estimate the reserves in the analytical block being
identified (Table 1, column 5).
The coefficient also takes part in the procedure of control sampling, where it regulates the number of the
obtained control samples (Table 1, column 6) (Nieć M. 1982).
Volver al índice

Depending on the value of the coefficient of variation, the minimal size of the sample may be determined, using
H. Czeczott's formula (formula 5)1 in which the coefficient "k" is assumed based on "V" (Mucha J., 1994). This
is due to the fact that for each mixture of borings, its smallest amount, which will not deviate from the
composition of the total primary sample, may be obtained. The size is mainly dependent upon the diameter of the
borings - the thicker is the sample material, the higher its heterogeneity is.

0 = k * d2 [kg] (5)

where:
k - the coefficient determined experimentally, characterizing the properties of the material relating to the
heterogeneity of mineralization and the diversification of physical properties of the individual components
(values in Table 1, column 7) - it is of estimative nature and allows for a large freedom of its selection,
d - the diameter of the largest borings in the sample, mm.

In ore deposits, the following types of mineralization may be distinguished based upon the coefficient of
variation: very uniform, uniform, non-uniform, very non-uniform and extremely non-uniform (Table 1, column
8).
This, among other things, allows to determine the samples of excessive mineralization, depending on the
variation of the "V" coefficient (Table 1, column 9). In case of identifying extremely high content, it must
always be checked if no mistake has occurred during the laboratory tests. If the sample preparation, the analysis
and the preparation of results cause no reservations, the correctness of the sample acquiring method must be
verified and it should be checked if it hadn't been acquired in a place of particularly high mineralization. It is
also advisable to acquire a second, control sample, at that same point. If the results of the laboratory analysis of
the control sample are comparable to the ones usually found in the deposit, the sample should be taken into
consideration in the calculation of the mean content of the useful component, while the previous sample is to be
omitted. In cases where the results of the analysis of the control sample show high values again, the range should
be determined. Most commonly, the quality of such part of deposit is appraised separately, especially when the
range of such part is considerable (Nieć M. 1982).
The distinguished types of mineralization also allow to determine the mass and the number of spot samples
acquired from the working using grid system (Table 1, columns 10 and 11).

Figure 31 -The diagrams of spot sampling in a grid system

1
Czeczott's formula does not provide information on the precision of the component content in the reduced sample.
Volver al índice

2. THE METHOD
The research concerning the application of the coefficient of variation have been performed in the conditions of
multistratum deposits of hard coal in the Upper Silesian Coal Basin (Fig. 2).

Figure 2 - The location of the research area

A Westphalian B stratum of power coal has been selected. The stratum is characterized by a simple geological
structure - it is low-faulted, with a small inclination angle and relatively simple and well-known tectonics.
The seam has been identified with 149 channel and core samples from boreholes, in which the basic chemical
and technological indicators have been denoted: the moisture content Wa, the ash contend Ad, the volatile matter
content Vdaf, the total sulphur Std and the calorific value CVdaf.
Based on this information, a tabular database concerning the coal quality, necessary for the further research, has
been created.
The compiled results have undergone statistical analysis, and basic statistical parameters have been specified for
each of the indicators, especially the "V" coefficient of variation.
The obtained results have been compared with the contents of the Table 1.
Volver al índice

3. THE RESULTS
The results of the statistical analysis which allow for the determination of the basic statistical parameters of the
coal quality indicators have been presented in Table 2.

Table 2 - The values of the selected statistical parameters of the analyzed quality indicators in the coal stratum
under research
Parameter Symbol Wa, % Ad, % Vdaf, % CVdaf, MJ/kg Std, %
Number of observations (samples) n 142 149 138 148 149
The highest (maximal) value max 6.8 50.6 42.5 34.6 2.1
The lowest (minimal) value min 0.8 3.1 22.7 26.7 0.2
Arithmetical mean X 4.0 12.4 33.1 31.5 0.9
Standard deviation s 1.2 7.2 4.7 0.8 0.3
Coefficient of variation V 31 58 14 3 36

The results compiled in Table 2 allow to classify the indicators of quality which differ in the values of the "V"
coefficient of variation into three groups of variation, based on the Baryszew classification (Table 1). The first
group (of low variation, with values of V=0-20%) includes the calorific value CVdaf (V=3%) and the volatile
matter content Vdaf (V=14%). The second group (of average variation, with values of V=20-40%) includes the
moisture content W a (V=31%) and the content of the total sulphur S t d (V=36%), that is the indicators which vary
in a relatively small range in the stratum. The third group (of high variation, with values of V=40-100%)
includes the ash content A d - usually the indicator which is hardest to characterize with respect to its variation.
None of the indicators has been assigned to the fourth and the fifth group, that is of very high (V=100-150%) and
extremely high (V>150%) variation.
These results translate into values specifying the "e%" precision of identification with a given probability, in line
with formula 2 (Table 3).

Table 3 - The "e%" precision of identification for the given confidence levels
Confidence level z Wa, % Ad, % Vdaf, % CVdaf, MJ/kg Std, %
68.3 % 1 2.5 4.8 1.1 0.2 2.9
95.5 % 2 5.1 9.5 2.3 0.5 5.9
99.7 % 3 7.6 14.3 3.4 0.7 8.8

The determined values of the "V" variation coefficient allow also to obtain initial information concerning the
proposed distances between the subsequent "l" points of sampling, which depend on the obtained "V" value, in
line with formula 4 (Table 4).

Table 4 - The distance between sampling points "l" [m] calculated using formula 4
Parameter Wa Ad Vdaf CVdaf Std
Coefficient of variation V, % 31 58 14 3 36
Sampling distance l, m 10 5 21 100 8

As the data presented in Table 4 shows, the calculated sampling distances are different for each parameter and
vary significantly, even by an order of magnitude.
The determined values of the "V" variation coefficient allow for a statement that 10 observations (samples) are
necessary to estimate reserves in an analytical block, while in control sampling 40 samples have to be acquired.
Volver al índice

4. CONCLUSIONS
The obtained results confirm that the numerically expressed variation of deposit parameters may constitute a
help in the solution of problems which are conditional to the identification and exploitation of a deposit, that is:
the appraisal of the coal quality in the deposit, the determination and design of an optimal grid for sampling and
identification of the deposit, the design of an optimal exploitation direction and the output control.
The results obtained by means of calculation may however be difficult or even impossible to apply in the
conditions of operation of coal mine plants.
While considering the specifics of the sampling complexity, the ranges in which the quality parameters of coal
vary and the "l" sampling distance possible to determine, it is proposed to apply a correcting coefficient to the
results presented in Table 4. The simplest solution (modification), which gives consideration to the specifics of
the mining plant operation is the introduction of a multiplier amounting to 10 to formula 4 (formula 6).

i 300 1Ar 1
l = *10|mJ
V
(6)

The multiplier will allow to relate the obtained values to the conditions of the Upper Silesian hard coal deposits
(Table 5). While considering the advised sampling distance for the Upper Silesian multistratum hard coal
deposits amounting to 50 m (Nieć M. 1982), the results should be rounded to multiplicities of 50 m (Marcisz M.,
2010, 2012).

Table 5 - The calculated distance between the "l" [m] sampling points
Parameter Wa Ad Vdaf CVdaf Std
Coefficient of variation V, % 31 58 14 3 36
Sampling distance l, m 100 50 210 1000 80
Sampling distance l after the rounding to the
100 50 200 1000 100
multiplicities of 50, m

REFERENCES
Marcisz M. (2012). Rola i znaczenie współczynnika zmienności w geologii górniczej i złożowej, [in „Wybrane
problemy badań geologicznych i hydrogeologicznych dla górnictwa i energetyki, Bukowski P. (ed.)],
Published by GIG.
Marcisz M. (2010). Szacowanie gęstości opróbowania pokładów węgla kamiennego w Górnośląskim Zagłębiu
Węglowym. Gliwice, Published by the Silesian University of Technology.
Mucha J. (1994). Metody geostatystyczne w dokumentowaniu złóż. Cracow, Published by the AGH University
of Science and Technology.
Nieć M. (1982). Geologia kopalniana. Warsaw, Wyd. Geologiczne.
Volver al índice

Table 1 - The classification of the deposit variation acc. to Baryszew, extended (Mucha J. 1994, Niec M.
1982)
The
Minimal Mass of
deposit
number of a single Number
variation The deposit The value
V observations Abnormally spot of spot
group variation Advised ofthe k
Coefficient required to high sample samples
Parameter based group No. of coefficient
of estimate the Mineralization content of acquired acquired
variation upon the (as provided control in H.
variation, reserves in the useful in the in the
V by descriptive samples Czeczott's
% an component grid grid
coefficient classification)2 formula
analytical system, system
of
block kg
variation
1 2 3 4 5 6 7 8 9 10 11
2-3 times
0.05 -
0-20 low 1 I 5 very uniform higher than - -
0.10
mean
15-20
4-5 times
0.10 -
20-40 average 2 II - uniform higher than 0.25 12-16
0.20
mean
8-10 times
0.20 -
40-100 high 3 8-10 30-40 non-uniform higher than 0.50 20-25
0.40
mean
10-15 times
0.40 - very non-
100-150 very high 4 III - higher than - -
0.80 uniform
mean
50-60
>15 times
extremely 0.80 - extremely non-
>150 5 15-25 higher than 1.00 35-50
high 1.00 uniform
mean

2
The characteristics of the deposit variation (or a part of such) is performed descriptively by distinguishing - depending on the
geological structural complexity - three groups: I -deposits of simple geological structure, non-faulted or low-faulted, with a small
thickness and mineral quality variation (considerably exceeding the balance boundary) in the deposit; II - deposits of mixed
geological structure, faulted, with varying thickness of the useful material and high variation of the main useful components in the
deposit; III - deposits of very varying and complex geological structure, non-continuous, faulted, characterized by very high
variation of thickness and quality of the useful material (Niec M. 1982).

View publication stats

You might also like