# RECORD 416244dcac790a50e1100858c6071c54dd1650ba
# TITLE: Analysing location attributes with a hedonic model for apartment prices in Donetsk, Ukraine
# YEAR: 2007.0  VENUE: nan  DOI: 10.1080/1648715x.2007.9637567
# AUTHORS: Marko Kryvobokov; Mats Wilhelmsson
# PAGES: 22  CANDIDATE TABLE PAGES (1-based): [10, 13, 21, 22]
# ABSTRACT: A hedonic model is specified for asking prices for apartments in Donetsk (Ukraine). This model is used to determine statistically significant location attributes. These attributes can be used for land assessment in a city where data on the land market are lacking. Distance gradients for CBD accessibility are investigated in different geographical directions. Separate models are created for sub‐samples located inside and outside the city centre. A spatial weight matrix is used to detect spatial autocorrelation. The regression results are compared with the valuation of experts.


===== PAGE 1 (score 5, numeric-paren 0, table-words 0, var-words 5) =====

International Journal of Strategic Property Management
ISSN 1648-715X print / ISSN 1648-9179 online © 2007 Vilnius Gediminas Technical University
http://www.ijspm.vgtu.lt
International Journal of Strategic Property Management (2007) 11, 157–178
ANALYSING LOCATION ATTRIBUTES WITH A HEDONIC
MODEL FOR APARTMENT PRICES IN DONETSK,
UKRAINE
Marko KRYVOBOKOV 1  and Mats WILHELMSSON 2
1 Royal Institute of Technology, Section of Real Estate Planning and Land Law, Stockholm,
Sweden and Donetsk National Technical University, Section of Geoinformatics and Geodesy,
Artyoma Str. 58, 83000 Donetsk, Ukraine
E-mail: k_mark@ukr.net
2 Royal Institute of Technology, Section of Building and Real Estate Economics, Brinellvägen 1,
S-100 44 Stockholm, Sweden
E-mail: matsw@infra.kth.se
Received 13 December 2006; accepted 22 June 2007
ABSTRACT. A hedonic model is specified for asking prices for apartments in Donetsk (Ukraine).
This model is used to determine statistically significant location attributes. These attributes
can be used for land assessment in a city where data on the land market are lacking. Distance
gradients for CBD accessibility are investigated in different geographical directions. Separate
models are created for sub-samples located inside and outside the city centre. A spatial weight
matrix is used to detect spatial autocorrelation. The regression results are compared with
the valuation of experts.
KEYWORDS: Location attributes; Land assessment; Apartment prices; Regression; Donetsk
1. INTRODUCTION
In Ukraine, land, whether urban or rural,
is taxed, while buildings and structures are
not. However, in most Ukrainian cities the
market for land is underdeveloped. There are
few land sales, and many of them involve land
sold by the city to the private sector. For ex-
ample, according to Thomas (2003), there is
effectively no secondary market for land in
Kiev. Therefore, the assessed value of urban
land in Ukraine is not based on market prin-
ciples (Thomas, 2003). Instead, according to
“Pattern of technical documentation of mon-
etary valuation of land” (1998), assessed value
depends on a weighted average of several lo-
cation attributes. Valuers, land managers, and
urban planners choose these location attributes
and weights of their relative importance based
on subjective judgements.
The process of property assessment can be
recognised as “quite subjective” even in coun-
tries with a long history of property taxation
(Cornia and Slide, 2005). The problem of sub-
jectivity in land assessment in Ukraine is de-
scribed in detail in Kryvobokov (2004). To in-
crease the degree of objectivity in the specifi-
cation of a model, he proposes that a list of
the most important location attributes that
influence values in well-developed foreign real


===== PAGE 10 (score 17, numeric-paren 6, table-words 4, var-words 3) =====

166
M. Kryvobokov and M. Wilhelmsson
apartment attributes and all location at-
tributes, denoted LIN (the linear model) and
LOG (the log-linear model) in Table 3. Model
LOG has higher adjusted R2 than model LIN
(77 percent compared to 69 percent). In both
models, Build, Area, Cond6, Cond7, DCBD,
and Prest are statistically significant. Cond5
is significant in model LIN, whereas FloorT
and Phone are significant in model LOG.
Where the location attributes are concerned,
DWater and DNuis are significant only in
model LIN, whereas DSCent and DRail are sig-
nificant only in model LOG. Among significant
location attributes measured in kilometers in
model LIN, the highest coefficients are for
DNuis and DWater; the lowest is that for
DCBD. At the same time the highest of the
significant location coefficients in model LOG
is that for DCBD.
To investigate the existence of multicolli-
nearity, we estimated the maximum of vari-
ance inflationary factors (VIF) (Table 3). The
principle that VIF in excess of 10 (or tolerance
lower than 0.10) indicates multicollinearity is
usually used in the literature, see e.g. Seiler
et al. (2001) and Des Rosiers et al. (2001). For
models LIN and LOG, the magnitudes of VIF
are lower than the threshold that indicates no
problem with multicollinearity.
Importantly, the values of the Moran’s I sta-
Table 3. Estimated OLS results for the whole data set
le
d
o
m
d
n
a
s
elb
air
a
V
s
citsir
e
tc
a
r
a
h
c
le
d
o
M
N
I
L
G
O
L
R
I
D
N
I
L
R
I
D
G
O
L
t
n
a
ts
n
o
C
5
5.0
6
0
3
)
3
9.0
(
7
9.7
*
)
3
1.2
2
(
5
9.6
2
2
5
)
5
5.1
(
*
)
8
0.2
2
(
1
9.7
1
r
o
ol
F
2
3.4
4
4
1
-
)
9
6.1
-(
2
0.0
-
)
0
7.0
-(
7
3.8
9
4
1
-
)
0
8.1
-(
)
2
6.0
-(
2
0.0
-
T
r
o
ol
F
0
3.0
7
5
1
-
)
7
7.1
-(
6
0.0
-
*
)
7
2.2
-(
0
6.1
5
8
1
-
*
)
5
1.2
-(
*
)
3
1.2
-(
6
0.0
-
dli
u
B
4
0.4
8
2
3
*
)
5
7.4
(
5
1.0
*
)
3
5.6
(
0
5.5
1
1
3
*
)
4
6.4
(
*
)
8
4.6
(
5
1.0
e
n
o
h
P
6
9.1
1
3
1
)
5
5.1
(
8
0.0
*
)
7
8.2
(
7
9.8
9
0
1
)
5
3.1
(
*
)
1
6.2
(
7
0.0
a
er
A
4
4.8
7
4
*
)
9
5.7
(
5
6.0
*
)
4
2.7
(
0
5.1
6
4
*
)
4
2.7
(
*
)
6
4.7
(
7
6.0
cla
B
3
5.8
2
5
-
)
2
5.0
-(
1
0.0
-
)
4
3.0
-(
8
6.5
1
9
-
)
2
9.0
-(
)
4
1.0
-(
7
0.0
-
2
d
n
o
C
6
0.2
1
3
1
)
7
1.1
(
2
0.0
)
0
6.0
(
2
1.4
1
6
1
)
6
4.1
(
)
4
6.0
(
2
0.0
3
d
n
o
C
8
6.7
6
9
1
)
3
4.1
(
2
0.0
)
2
4.0
(
6
2.0
9
9
1
)
3
4.1
(
)
3
3.0
(
2
0.0
4
d
n
o
C
0
7.6
2
0
2
)
5
7.1
(
4
0.0
)
1
1.1
(
0
5.4
0
2
2
)
4
9.1
(
)
5
2.1
(
4
0.0
5
d
n
o
C
3
5.6
6
2
2
*
)
0
1.2
(
5
0.0
)
8
5.1
(
5
3.1
2
3
2
*
)
7
1.2
(
)
8
7.1
(
6
0.0
6
d
n
o
C
9
1.0
0
6
4
*
)
6
7.2
(
3
1.0
*
)
0
5.2
(
5
7.5
1
6
4
*
)
9
7.2
(
*
)
5
7.2
(
4
1.0
7
d
n
o
C
8
4.3
1
7
9
*
)
8
1.6
(
0
3.0
*
)
8
0.6
(
8
3.1
0
4
9
*
)
9
7.5
(
*
)
6
3.6
(
1
3.0
)
d
e
u
nit
n
o
c
(


===== PAGE 11 (score 5, numeric-paren 1, table-words 1, var-words 2) =====

167
Analysing Location Attributes with a Hedonic Model for Apartment Prices in Donetsk, Ukraine
tistic are quite large; this suggests that nei-
ther model captures the underlying spatial re-
lationship very well. As an alternative, we con-
sider spatial models to incorporate the spatial
structure into the analysis (Anselin, 1988 is
the standard reference documenting these
models). We have estimated two different kinds
of spatial regression models to correct for the
underlying spatial relationship. These are a
spatial autoregressive model and a spatial er-
ror model. The results for the spatial models
are presented in the Appendix. Even though
the coefficient for spatial dependency is highly
significant, none of the estimates for the loca-
Note: White heteroskedastic consistent estimates. t-statistics are in parentheses below the coefficients. The asterisk
indicates significance at the 5% level.
le
d
o
m
d
n
a
s
elb
air
a
V
s
citsir
e
tc
a
r
a
h
c
le
d
o
M
N
I
L
G
O
L
R
I
D
N
I
L
R
I
D
G
O
L
)
d
e
u
nit
n
o
c
(
D
B
C
D
1
2.8
2
8
-
*
)
2
7.6
-(
4
2.0
-
*
)
4
2.0
1
-(
7
7.5
3
9
1
-
*
)
6
6.7
-(
*
)
2
8.6
-(
3
2.0
-
t
n
e
C
S
D
2
6.7
4
-
)
3
2.0
-(
5
0.0
-
*
)
1
9.4
-(
7
6.6
9
1
-
)
7
8.0
-(
*
)
7
4.3
-(
4
0.0
-
p
ot
S
D
0
5.4
6
8
2
-
)
6
3.1
-(
1
0.0
-
)
4
3.0
-(
4
9.5
7
9
5
-
*
)
0
9.2
-(
)
6
2.0
-(
0
0.0
-
lia
R
D
2
4.7
3
-
)
1
4.0
-(
1
1.0
-
*
)
4
4.5
-(
3
5.3
0
3
*
)
0
3.2
(
*
)
5
2.4
-(
1
1.0
-
reta
W
D
0
6.2
0
7
1
-
*
)
9
7.3
-(
1
0.0
-
)
7
3.0
-(
2
8.9
9
2
-
)
8
5.0
-(
)
3
9.0
-(
1
0.0
-
si
u
N
D
5
3.6
1
8
2
*
)
0
8.3
(
3
0.0
)
4
5.1
(
4
9.4
8
1
3
*
)
8
1.4
(
)
9
2.1
(
3
0.0
ts
er
P
7
5.1
6
7
4
*
)
7
9.5
(
9
0.0
*
)
4
7.2
(
8
1.3
5
4
2
*
)
7
7.2
(
*
)
1
0.2
(
8
0.0
D
B
C
D
·
N
-
-
7
4.7
3
7
*
)
1
9.2
(
)
5
3.0
(
1
0.0
D
B
C
D
·
S
-
-
6
7.1
6
-
)
2
5.0
-(
)
6
5.1
-(
3
0.0
-
D
B
C
D
·
W
-
-
2
6.6
4
7
*
)
1
0.5
(
)
4
7.1
-(
4
0.0
-
s
n
oit
a
v
r
e
s
b
O
5
2
3
5
2
3
5
2
3
5
2
3
2
R
d
e
ts
u
j
d
A
9
6.0
7
7.0
1
7.0
7
7.0
e
ula
v-
F
6
7.9
3
7
4.7
5
6
7.7
3
0
2.0
5
x
a
M
F
I
V
6.4
(
D
B
C
D
)
7.3
(
D
B
C
D
)
2.0
2
(
D
B
C
D
)
)
D
B
C
D
(
1.6
I
s'n
a
r
o
M
6
3.7
0
6.3
9
4.6
4
7.3
tion variables change dramatically. None of
them are significantly different from the re-
sults of the OLS model. Hence, we use the OLS
estimates in our econometric analysis.
As in Söderberg and Janssen (2001), we
analyse whether the CBD accessibility affects
the asking price differently in different geo-
graphical directions. We divide Donetsk into
four parts, North, East, South, and West, with
the CBD as the origin, to carry this out. Con-
sidering that the area to the east of the CBD
is the most prestigious area outside the city
centre, area East is used as the default area.
So, we introduce three dummies for location


===== PAGE 13 (score 13, numeric-paren 11, table-words 1, var-words 0) =====

169
Analysing Location Attributes with a Hedonic Model for Apartment Prices in Donetsk, Ukraine
le
d
o
m
d
n
a
s
elb
air
a
V
s
citsir
e
tc
a
r
a
h
c
le
d
o
M
N
I
L
G
O
L
R
I
D
N
I
L
R
I
D
G
O
L
t
n
a
ts
n
o
C
9
3.9
4
7
1
6
*
)
1
2.2
(
5
0.9
*
)
7
0.7
(
9
5.4
1
5
0
1
*
)
0
5.3
(
8
8.8
*
)
5
6.8
1
(
1
r
o
ol
F
4
6.9
3
7
7
-
*
)
8
2.2
-(
0
2.0
-
*
)
4
2.2
-(
0
1.0
4
2
-
)
3
3.0
-(
1
0.0
-
)
2
2.0
-(
T
r
o
ol
F
4
3.4
3
9
2
-
)
8
4.1
-(
0
1.0
-
)
4
8.1
-(
0
3.7
8
9
-
)
8
3.1
-(
5
0.0
-
)
0
6.1
-(
dli
u
B
3
5.7
9
8
6
*
)
5
6.2
(
6
1.0
*
)
2
2.2
(
2
8.1
4
7
3
*
)
7
7.5
(
0
2.0
*
)
1
7.7
(
e
n
o
h
P
5
1.6
8
1
-
)
6
0.0
-(
4
0.0
)
0
5.0
(
9
8.8
9
9
)
0
5.1
(
6
0.0
*
)
8
1.2
(
a
er
A
2
0.6
6
5
*
)
0
8.4
(
6
8.0
*
)
1
8.4
(
3
1.3
6
3
*
)
7
0.6
(
3
4.0
*
)
3
5.3
(
cla
B
5
8.7
8
2
2
-
)
8
0.1
-(
5
0.0
-
)
0
8.0
-(
0
1.7
9
5
-
)
6
8.0
-(
0
0.0
-
)
8
0.0
-(
2
d
n
o
C
6
5.5
0
2
-
)
6
0.0
-(
1
0.0
-
)
1
1.0
-(
4
4.8
2
8
)
7
7.0
(
2
0.0
)
6
4.0
(
3
d
n
o
C
3
2.2
9
7
1
)
8
3.0
(
4
0.0
)
3
3.0
(
3
3.9
4
3
1
)
2
9.0
(
2
0.0
)
1
3.0
(
4
d
n
o
C
7
5.9
3
2
1
)
8
3.0
(
1
0.0
)
7
0.0
(
9
8.3
7
4
)
6
4.0
(
4
0.0
)
4
0.1
(
5
d
n
o
C
5
7.1
5
7
-
)
3
2.0
-(
4
0.0
-
)
3
4.0
-(
2
7.9
9
8
1
)
0
9.1
(
9
0.0
*
)
1
2.2
(
6
d
n
o
C
8
7.5
8
8
9
*
)
8
2.2
(
2
2.0
)
5
8.1
(
4
7.3
0
6
2
*
)
1
0.2
(
2
1.0
*
)
5
2.2
(
7
d
n
o
C
4
5.7
4
1
7
)
4
4.1
(
5
1.0
)
7
1.1
(
1
2.9
1
2
9
*
)
2
4.6
(
3
3.0
*
)
9
7.5
(
D
B
C
D
5
5.0
6
2
6
-
*
)
5
5.2
-(
3
1.0
-
*
)
6
0.2
-(
9
3.2
1
6
-
*
)
4
6.5
-(
1
3.0
-
*
)
5
6.0
1
-(
t
n
e
C
S
D
2
1.6
3
1
1
-
)
8
2.0
-(
2
1.0
-
)
4
4.0
-(
0
8.4
9
7
-
*
)
5
7.3
-(
4
0.0
-
*
)
6
4.3
-(
p
ot
S
D
9
8.3
2
5
3
)
0
4.0
(
1
0.0
)
9
0.0
(
3
4.7
5
7
2
)
8
5.1
(
1
0.0
)
4
5.0
(
lia
R
D
6
5.5
8
1
5
-
)
4
8.1
-(
3
8.0
-
)
8
4.1
-(
1
8.7
1
2
-
*
)
3
4.2
-(
8
0.0
-
*
)
5
9.3
-(
reta
W
D
4
6.7
1
0
8
)
0
0.1
(
7
0.0
)
3
9.0
(
0
2.2
4
4
1
-
*
)
8
2.3
-(
4
0.0
-
*
)
3
0.2
-(
si
u
N
D
6
1.9
3
8
4
-
)
3
8.0
-(
2
1.0
-
)
4
5.0
-(
1
5.4
2
7
)
9
9.0
(
3
0.0
)
2
7.1
(
ts
er
P
-
-
5
1.2
8
2
2
*
)
1
0.3
(
6
0.0
*
)
2
0.2
(
s
n
oit
a
v
r
e
s
b
O
5
7
5
7
0
5
2
0
5
2
R
d
e
ts
u
j
d
A
2
7
4.0
7
4.0
9
6.0
3
7.0
e
ula
v-
F
3
6.4
0
7.4
4
7.0
3
0
8.6
3
F
I
V
x
a
M
9.8
(
lia
R
D
)
0.9
(
lia
R
D
)
0.4
(
lia
R
D
)
4.3
(
5
d
n
o
C
)
Table 4. Estimated OLS results for split data sets
Note: In models LIN_Cent and LOG_Cent, Prest is equal to 1 for all apartments.


===== PAGE 14 (score 24, numeric-paren 0, table-words 10, var-words 4) =====

170
M. Kryvobokov and M. Wilhelmsson
respectively, both are larger than the critical
value of 1.57 at the 5% significance level.
WSE for sub-samples in total is calculated
using the formula
∑
∑
=
=
−
−
⋅
−
−
=
m
i
i
i
m
i
i
i
i
k
n
SE
k
n
WSE
1
1
)1
(
)
)1
((
,
where: n is the number of observations; k is
the number of explanatory variables; i is the
count for sub-samples; m is the number of sub-
samples; SEi is the standard error in the i-th
sub-sample model.
The results of WSE estimations are pre-
sented and compared with the initial models
in Table 5. For linear specifications, the reduc-
tion in the standard error is 17 percent,
whereas for log-linear specifications it is only
8 percent. According to literature review by
Watkins (2001), either 5 percent or 10 percent
cut off is employed without a strict guidance
on the size of the reduction. In our case, the
sub-market specifications in a linear form sat-
isfy both thresholds, while the log-linear mod-
els satisfy only the former one.
Thus, Chow-test and WSE estimation indi-
cate that two delineated sub-markets are dif-
ferent and cannot be described by one hedonic
model. There are some interesting findings
when it comes to the location variables. DCBD
is the only significant location variable in mod-
els LIN_Cent and LOG_Cent and the coefficient
for DSCent is smaller than that for DCBD. For
the linear model the former is 5.5 times smaller
than the latter. In models LIN_Out and
LOG_Out, all the location variables except
DStop and DNuis are significant. LIN_Out is
the only linear model where DSCent is signifi-
cant. In this model, the coefficient for DSCent
is higher than that for DCBD; the coefficient
for DWater is 2.4 times higher than that for
DCBD. The remarkable point is the compari-
son of the coefficients for DCBD in LIN_Cent
and LIN_Out. For central locations, it is more
than 10 times higher than that for locations
outside the centre. At the same time, in model
LOG_Out the coefficient for DCBD is the high-
est among coefficients for the location vari-
ables. For central locations, the insignificant
variables DWater and DNuis have unexpected
signs. This might be explained by the high den-
sity of buildings, which are barriers for the
influence of water, green areas and nuisance
objects.
Though there are dissimilarities between
the linear and log-linear models concerning the
importance of secondary centres, it seems that
these centres in Donetsk are important. There-
fore, Donetsk can be considered as a non-
monocentric city. The difference between non-
monocentric and polycentric urban models is
described e.g. in Sivitanidou (1997). Here we
do not investigate this difference, but conclude
that a monocentric model is not appropriate
for Donetsk.
Table 5. Weighted standard error tests
le
d
o
M
r
o
rr
e
d
r
a
d
n
a
t
S
n
oitc
u
d
e
r
t
n
e
c
r
e
P
:s
n
oit
a
cific
e
p
s
r
a
e
ni
L
N
I
L
9
7
4
5
t
u
O
_
N
I
L
d
n
a
t
n
e
C
_
N
I
L
8
3
5
4
7
1
:s
n
oit
a
cific
e
p
s
r
a
e
nil-
g
o
L
G
O
L
4
0
9
5
7
1.0
t
u
O
_
G
O
L
d
n
a
t
n
e
C
_
G
O
L
3
3
3
2
6
1.0
8


===== PAGE 21 (score 24, numeric-paren 24, table-words 0, var-words 0) =====

177
Analysing Location Attributes with a Hedonic Model for Apartment Prices in Donetsk, Ukraine
APPENDIX
Spatial Autoregressive Models (SAR) and Spatial Error Models (SEM)
d
n
a
s
elb
air
a
V
le
d
o
m
s
citsir
e
tc
a
r
a
h
c
le
d
o
M
G
O
L
R
I
D
G
O
L
R
A
S
M
E
S
R
A
S
M
E
S
t
n
a
ts
n
o
C
7
9.8
*
)
4
0.9
(
5
1.8
*
)
6
7.1
2
(
2
1.6
*
)
8
1.9
(
9
0.8
*
)
4
7.1
2
(
1
r
o
ol
F
2
0.0
-
)
2
8.0
-(
3
0.0
-
)
9
9.0
-(
2
0.0
-
)
7
7.0
-(
2
0.0
-
)
0
9.0
-(
T
r
o
ol
F
6
0.0
-
*
)
1
5.2
-(
6
0.0
-
*
)
3
3.2
-(
6
0.0
-
*
)
1
4.2
-(
5
0.0
-
*
)
4
2.2
-(
dli
u
B
4
1.0
*
)
1
1.6
(
6
1.0
*
)
9
5.6
(
4
1.0
*
)
4
1.6
(
6
1.0
*
)
8
5.6
(
e
n
o
h
P
7
0.0
*
)
7
7.2
(
6
0.0
*
)
4
5.2
(
7
0.0
*
)
7
5.2
(
6
0.0
*
)
3
3.2
(
a
er
A
0
6.0
*
)
0
4.6
(
0
6.0
*
)
5
2.6
(
2
6.0
*
)
6
5.6
(
1
6.0
*
)
4
4.6
(
cla
B
1
0.0
-
)
1
3.0
-(
1
0.0
-
)
8
0.0
-(
1
0.0
-
)
4
1.0
-(
1
0.0
)
1
1.0
(
2
d
n
o
C
1
0.0
)
4
3.0
(
1
0.0
)
6
2.0
(
1
0.0
)
8
3.0
(
1
0.0
)
0
3.0
(
3
d
n
o
C
2
0.0
)
9
2.0
(
1
0.0
)
1
2.0
(
1
0.0
)
5
2.0
(
1
0.0
)
6
1.0
(
4
d
n
o
C
3
0.0
)
2
8.0
(
3
0.0
)
1
7.0
(
3
0.0
)
1
9.0
(
3
0.0
)
0
8.0
(
5
d
n
o
C
5
0.0
)
1
3.1
(
5
0.0
)
3
3.1
(
5
0.0
)
4
4.1
(
5
0.0
)
6
4.1
(
6
d
n
o
C
2
1.0
*
)
6
5.2
(
3
1.0
*
)
5
6.2
(
3
1.0
*
)
0
8.2
(
4
1.0
*
)
8
8.2
(
7
d
n
o
C
0
3.0
*
)
2
9.5
(
0
3.0
*
)
3
8.5
(
1
3.0
*
)
1
0.6
(
0
3.0
*
)
4
9.5
(
D
B
C
D
9
1.0
-
*
)
1
1.9
-(
3
2.0
-
*
)
8
1.1
1
-(
8
1.0
-
*
)
5
2.7
-(
1
2.0
-
*
)
2
7.7
-(
t
n
e
C
S
D
5
0.0
-
*
)
6
2.4
-(
5
0.0
-
*
)
4
0.4
-(
4
0.0
-
*
)
6
2.3
-(
4
0.0
-
*
)
4
8.2
-(
p
ot
S
D
1
0.0
-
)
2
3.0
-(
1
0.0
-
)
2
6.0
-(
1
0.0
-
)
2
2.0
-(
1
0.0
-
)
2
5.0
-(
lia
R
D
8
0.0
-
*
)
2
3.4
-(
1
1.0
-
*
)
6
8.4
-(
8
0.0
-
*
)
6
5.3
-(
0
1.0
-
*
)
1
0.4
-(
reta
W
D
1
0.0
-
)
6
3.0
-(
1
0.0
-
)
8
6.0
-(
1
0.0
-
)
7
8.0
-(
2
0.0
-
)
7
2.1
-(
)
d
e
u
nit
n
o
c
(


===== PAGE 22 (score 10, numeric-paren 8, table-words 1, var-words 0) =====

178
M. Kryvobokov and M. Wilhelmsson
d
n
a
s
elb
air
a
V
le
d
o
m
s
citsir
e
tc
a
r
a
h
c
le
d
o
M
G
O
L
R
I
D
G
O
L
R
A
S
M
E
S
R
A
S
M
E
S
)
d
e
u
nit
n
o
c
(
si
u
N
D
2
0.0
)
6
1.1
(
2
0.0
)
8
1.1
(
2
0.0
)
0
0.1
(
2
0.0
)
1
9.0
(
ts
er
P
9
0.0
*
)
7
8.2
(
0
1.0
*
)
2
8.2
(
7
0.0
*
)
7
1.2
(
8
0.0
*
)
9
0.2
(
D
B
C
D
·
N
-
-
1
0.0
)
5
3.0
(
1
0.0
)
2
1.0
(
D
B
C
D
·
S
-
-
2
0.0
-
)
0
0.1
-(
4
0.0
-
)
5
5.1
-(
D
B
C
D
·
W
-
-
3
0.0
-
)
0
6.1
-(
4
0.0
-
*
)
7
0.2
-(
ρ
1
2.0
*
)
4
7.3
(
-
9
1.0
*
)
1
3.3
(
-
λ
-
2
2.0
*
)
2
3.3
(
-
1
2.0
*
)
8
1.3
(
s
n
oit
a
v
r
e
s
b
O
5
2
3
5
2
3
5
2
3
5
2
3
d
e
ts
u
j
d
A
R2
7
7.0
8
7.0
7
7.0
8
7.0