25 6. Counting ffss, the number of sub districts as subordinate status with
respect to every sub district. ffaa, the number of sub districts as ascribe advantage over every sub district, and ffii, the number of sub districts as
indefinite to every sub district.
7. Obtaining AA = 100 × ffaaDD, SS = 100 × ffssDD, II = 100 ×
ff
iiDD clearly, AA + SS + II = 100. 8. CCC = 100
– AA and 9. ccc is obtained by rounding CCC to two decimal places and then
multiplying by 100.
10. bbb = SSCCC, and imposing 0.999 as an upper limit. 11. Adding these two values, ccc and bbb as ccc.bbb.
12. Ranking of ccc.bbb values of all individuals is as salient. 13. Precedence plot graphing. Precedence plot is a scatter plot of SS as DD
and AA as DD. Dots represent the position of sub districts at the
ranking. North West part is the position of top level and South East is the bottom level.
Actually there are some other steps to continue, but not written here due to only two indicators in the application.
2.4. Results and Discussion
Paired scatter plots are a simple way to see and get the information about strength of the relationship between indicators. Scatter plots of indicators HIP and
PL is shown by Figure 9. Coefficient correlation between HIP and PL is 0.3543. It is believed, there is a positive statistical association p-value 2.2e-16 between
these two indicators. It means these two indicators have the same positive direction to measure poverty as an abstract concept. Figure 9 presents an
interesting distribution of the data, where a part of observations is on a linear line, while the other observations almost accumulated on the lower left part. The
observations on the linear line mean all poor families in those sub districts have the health insurance for the poor. The observations above the linear line mean not
all poor families in those sub districts have health insurance. Furthermore, the observations under the linear line show that in those sub districts, not all owners of
26 the health insurance for the poor are poor families. Numerical variability of
poverty indicators is shown by Figure 10. It shows two boxplots of two indicators, HIP and PL. This figure shows that HIP is more variable than PL. It seems that
HIP contains outliers.
HIP Figure 9 Scatterplot of indicators HIP vs PL HIP : number of health insurances
for the poor, PL = number of poverty letters
Figure 10 Boxplot of the indicators, HIP and PL Table 4 is six leading lines of identity number id, province, district, sub
district, indicators values, and sub district rank of every indicator in the data sets.
5000 10000
15000 20000
25000 5000
10000 15000
1 2
5000 15000
25000
PL
HIP PL
27 Seventh and 8
th
column are the place rank of HIP and PL, respectively, as the
result of R function called PlacRank in Schema 1 Myers and Patil 2010.
Table 4 The first 6 lines of the data: identity number id, province, district, sub district, indicators values, and sub district
’s ranking based on indicator
id Province
District Sub-district Indicator value
Indicator ranking HIP
PL HIP
PL 1
2 3
4 5
6 7
8 1
West Java Bogor
Nanggung 3402
634 1055
1004 2
West Java Bogor
Leuwiliang 7727
636 215 1002.5
3 West Java
Bogor Leuwisadeng
3131 976
1121.5 791
4 West Java
Bogor Pamijahan
7382 1275
245 683
5 West Java
Bogor Cibungbulang
9919 1423
82 646
6 West Java
Bogor Ciampea
6485 2084
344 537
… …
… ….
… …
… …
Rank numbers at 7
th
and 8
th
columns of Table 4 are used to place every sub districts at a particular position in Hasse diagram as entities. The Hasse diagram is
not shown because of too many entities 1679 in this computation.
From Hasse diagram, ffaa, ffss, and ffii are counted to obtained AA, SS, II
, CCC, ccc, bbb, and finaly ccc.bbb as ORDIT scores which will be ranked as
salient of sub district. Table 5 shows the head first 6 lines of the data frame obtained by applying
the ProdOrdr function to the place ranks of poverty measurement. In the current content, the idea of ‘better’ from reference Myers and Patil 2010 is replaced by
‘more severe’.
Table 5 The first 6 lines of the result obtained by applying the ProdOrdr function to place poverty measurement rank of sub district
id
Sub districts
AA SS
ORDIT Salnt
1
Nanggung
20.32 42.91 7968.539
994 2
Leuwiliang
38.5 10.91 6150.177
599 3
Leuwisadeng
21.57 35.46 7843.452
961 4
Pamijahan
53.22 8.4
4678.18 349
5
Cibungbulang
60.01 3.28 3999.082
258 6
Ciampea
55.6 7.99
4440.18 314
28 Precedence plot of sub districts by product-order is shown by Figure 11. The
dots represent 1679 sub districts, where the North West represents the most severe sub districts and the South East represents the least severe sub districts. Actually,
if the number of instances or sub district is not too big, the identity number of sub districts can be shown at the graph.
Figure 11 Precedence plot based on place ranks of sub districts from R commands
Sub districts listed at Table 6 and 7 are obtained from the North West of Figure 11. These sub districts are the top level and the bottom level of the ranking
result, where it means the most severe and the least severe in poverty. Table 6 and Table 7 show the ten most severe sub districts and the ten least
severe sub districts, respectively. The columns of the tables are case id, the name
of districts, the name of sub districts, ascribed advantage AA, subordinated status SS, ORDIT score, and salient or rank of the sub districts. Five of the ten most
severe sub districts are located in Jember district, and the rest are located in Karawang, Banyuwangi, and Bondowoso districts.
Five of the ten least severe sub districts at Table 7 are located in Probolinggo city, two of them are in Sampang district, and the rest are in Surabaya district.
Table 6 The ten most severe sub districts according to ORDIT ranking
20 40
60 80
100 20
40 60
80 100
SS as DD Deleted Domain A
A a
s D
D
29
No. id
District Sub district
AA SS
ORDIT Salnt
1 1320
Jember Kalisat
99.28 72
1 2
1305 Jember
Mumbulsari 98.99
0.06 101.059 2
3 1315
Jember BangsalSari
98.93 0.12 107.112
3 4
1313 Jember
SumberBaru 98.45
0.18 155.116 4
5 1303
Jember Silo
98.39 0.24 161.149
5 6
465 Karawang
Karawang Barat 98.21
0.3 179.168 6
7 1341
Banyuwangi Rogojampi
98.15 0.3 185.162
7 8
1322 Jember
SumberJambe 97.85
0.42 215.195 8
9 1333
Banyuwangi Muncar
97.79 0.48 221.217
9 10
1357 Bondowoso
Tlogosari 97.44
0.54 256.211 10
… …
… …
… …
… …
Table 7 The ten least severe sub districts according to ORDIT ranking
No. id
District Sub district
AA SS
ORDIT Salnt
… …
… …
… …
… …
1670 1594
Sampang Banyuates
0 96.84 10000.97 1652.5 1671
1595 Sampang
Robatal 0 96.84 10000.97 1652.5
1672 1635
Kota Probolinggo Kademangan A 0 96.84 10000.97 1652.5
1673 1636
Kota Probolinggo Kademangan B 0 96.84 10000.97 1652.5
1674 1637
Kota Probolinggo Wonoasih 0 96.84 10000.97 1652.5
1675 1638
Kota Probolinggo Mayangan A 0 96.84 10000.97 1652.5
1676 1639
Kota Probolinggo Mayangan B 0 96.84 10000.97 1652.5
1677 1672
Surabaya Pabean Cantian
0 96.84 10000.97 1652.5 1678
1674 Surabaya
Asemrowo 0 96.84 10000.97 1652.5
1679 1676
Surabaya Pakal
0 96.84 10000.97 1652.5
To whole results of grouping rank are summarized at table 8. According to the classification result on the table, the highest percentage in West Java is mild
level, 43.4 256 sub districts, whereas in Central Java is moderate level, 38.8 216 sub districts, and in East Java is severe level, 53.1 283 sub districts.
Based on this result, it is believed the order of these three provinces from severe to mild is East Java, Central Java, and West Java.
30 Table 8 Poverty level of sub districts in the West, Central and East Java
Province Poverty level of sub-districts
Total sub districts
worst moderate
mild West
Java Count
142 192
256
590 within province
24.1 32.5
43.4
100.0 Central
Java Count
133
216
207 556
within province 23.9
38.8
37.2 100.0
East Java Count
283
150 100
533 within province
53.1
28.1 18.8
100.0 Total
Count 558
558 563
1679 within province
33.2 33.2
33.5 100.0
2.5 Conclusion
According to the ORDIT analysis with two indicators, i.e. the number of health insurances for the poor and the number of poverty letters, 6 of the most
severe sub-districts of poverty are located in Jember district. This result is appropriate with the result of other poverty researches and the news. According to
the news, Sumberjambe, a most severe sub district in this result, is a contributor to poverty in East Java. Based on the Data Collection Program of Social Protection in
2008, Sumberjambe has 63 or 12,827 poor households Sumberjambedesa 2010. On the other hand, 5 and 3 of the least severe are located in Probolinggo
city and Surabaya, respectively. The least severe sub districts are located in cities area that easier to get education, information, and health center facilities.
Based on the information from Research and Development in East Java Province, Sampang is one of worst area in poverty. About 60 of population is the
poor. But the population size is small and the total number of the poor is not as many as the most severe sub district BPPPT 2010. So it seems right to include
this sub district as the least severe level. Related to the rigorous efforts through the instruction set to handle the
underdeveloped villages, this result gives input and advantage to the government in decision making of sub districts priority determining to take action in poverty
reducing. Increasing primary health care is an aptly action according to a conclusion of World Bank research in health sector, whic
h concluded “the greatest
31 benefit to the poor would come from an increase in primary
health care spending” Lanjouw 2001.
According to the ranking result, sequence of poverty level in those 3 provinces in this study, from worst to mild, is East Java, Central Java, and West
Java. ORDIT as a ranking method is convenient to obtain ranking for big number of observations. As stated before, this study using 1679 sub districts to be ranked,
and this number is not few, but ORDIT give acceptable result, where it is appropriate to an article news that said, “Three provinces, Central Java,
Yogyakarta, East Java has the highest poverty rates” Susanto and Darmawan
2010.
32
33
Chapter 3 COMPARISON BETWEEN CIRCLE BASED SCAN
STATISTICS AND UPPER LEVEL SET SCAN STATISTICS BASED ON SIMULATION STUDY
3.1 Introduction
“Hotspot” is an area that has an elevated response compared with the rest area. Detection or identification this area is important in many fields of study.
Spatial epidemiology is a field that needs hotspot detection method most to help identifying environmental factors, associated with disease or many other indicators
related to a measurement of a concept in a region. There are some hotspot detection methods developed by experts, some of
which are: 1 a spatial scan statistic with SaTScan software Kulldorff 1997, 2 Upper Level Set scan statistic Patil and Taillie 2004, 3 generalized additive
models GAM Hastie and Tibshirani 1991, and 4 Bayesian disease mapping BYM Besag et al. 1991. The hotspot detection methods 1 and 2 have the
same basic theory where the difference is in the way they do the scanning process. Circle-based Scan Statistic is a hotspot detection method based on circle shape
detection area. This method builds circle continuously wider to detect hotspot area. Whereas, Upper Level Set scan statistic builds adjacent cells with highest intensity
to find hotspot area. This chapter focuses on these two hotspot detection methods. The objective of this chapter is to compare two hotspot detection methods
suitable for detecting local spatial clusters, i.e. Circle-based Scan Statistic SS and Upper Level Set scan statistic ULS with the observed incidence is assumed as
Poisson distribution. For the next explanation, the terms SS and ULS are used for Circle-based Scan Statistic and Upper Level Set scan statistic, respectively. A
simulation is carried out to compare these two methods. A geographic cluster is chosen as high-risk areas. For each simulation, the performance of the methods is
assessed in terms of the sensitivity, specificity, and percentage correctly classified for each cluster. As the detail, performance is measured through 14 criteria.