You are on page 1of 16

REGRESSION ANALYSYS: AN EXAMPLE OF APPLICATION

The present study investigated the relationship between location and quality of life in
informal settlements in the case of Tacloban City. This chapter discusses the
determinants of and factors affecting location of informal settlements based on the
results of logistic regression analysis and multiple regression analysis. The survey
data on sample informal settlement households was used for analysis. In logistics
regression analysis, location in danger and non-danger area is considered as response
variable and out of variables as potential explanatory variables, six variables -
Distance from the CD, !verage "onthly Transport Cost, #and $wnership, %ouse
Type, Toilet Type and #and &se predict location. In multiple regression analysis,
monthly household income was considered as dependent variable and out of
independent variables including location six variables ' %ousehold (i)e, *umber of
Income earners, %ousehold "onthly +xpenditure, %ousehold "onthly Clothing
+xpenditure explain the household income.
Logistic Regression Analysis
In order to ascertain the lin,ages between location and quality of life in terms of
income, housing and basic services li,e piped water, sanitation, electricity, garbage
disposal, drainage, paved roads in informal settlements in Tacloban City, the statistical
tool ' logistic regression analysis was used ta,ing #$C!TI$*, as the dichotomous
response variable assuming two values- . represents D!*/+0 !0+! and 1
represents *$*-D!*/+0 !0+! and 2 metric and 3 dummy coded 4categorical5non-
metric6 variables as explanatory variables. In this analytical predictive technique -
multiple 4two-group6 discriminant analysis, it was purposed to determine if
.
statistically significant differences exist between informal settlements located in
danger area and non-danger area and which independent variables account most for
the differences.
The mean of the dichotomous random variable, designated, as p is the proportion of
times that it ta,es the value of ..
p 7 849 7 .6
7 8 4Danger !rea6
The logistic function for this regression model is
p 7
e
a : b
.
;
.
: b
<
;
<
: = : b
n
;
n
>here p 7 the dichotomous variable - #ocation 4. 7 Danger !rea? 1 7 *on-Danger
area, a 7 the intercept, b
i
7 the coefficients of regression and ;
i
7 the explanatory
variables.
This is not the standard regression form -
9 7 a : b
.
;
.
:
<
;
<
: = : b
n
;
n
(ince, here p is a probability and therefore it is restricted to ta,e values between 1 and
..
The model was expressed in the form-
p 7
e
a : b
.
;
.
: b
<
;
<
: = : b
n
;
n
5 .:
e
a : b
.
;
.
: b
<
;
<
: = : b
n
;
n
<
To overcome certain constraints in the model so that the expression cannot yield a
value that is either negative or greater than ., thus it restricts the estimated values of p
to the required range.
If p 7
e
a : b
.
;
.
: b
<
;
<
: = : b
n
;
n
5 .:
e
a : b
.
;
.
: b
<
;
<
: = : b
n
;
n
Then, p5. ' p 7 p 7
e
a : b
.
;
.
: b
<
;
<
: = : b
n
;
n
5.5.:
e
a : b
.
;
.
: b
<
;
<
: = : b
n
;
n
The logistic regression expressed ta,ing the logarithm of each side,
ln @p5 .-pA 7 a : b
.
;
.
:
<
;
<
: = : b
n
;
n
Instead of assuming he relationship between p and ;
i
linear, the relationship between
ln @p54.-p6A and ;
i
is linear.
The logistic regression equation was done to model p, the probability that informal
settlements located in danger area
0esponse Bariable
#oc< 7 #ocation of the Informal (ettlement
. 7 Danger !rea
1 7 *on ' Danger !rea
C
8otential +xplanatory Bariables
DistC 7 Distance from Central usiness District5Downtown
%%(iD 7 %ousehold (i)e
0oom<2 7 *umber of 0ooms
Inc+CE 7 *umber of Income +arning "embers
%%IncC3 7 %ousehold Income
Tra>,FE 7 "onthly %ousehold Transport +xpenditure for >or,place
%%+xG1b 7 !verage "onthly %ousehold +xpenditure
#!mt2G 7 !mount of #and $ccupied
Du$c22 7 Duration of $ccupation
D#$c<D 7 #ot occupancy
1 7 /overnment
. 7 *on-/overnment
D%Typ<F 7 %ouse Type
1 7 Concrete
. 7 *ot Concrete 4>ooden5 #ight "aterials6
DToilTC.
1 7 $pen 8it 4$pen 8it5(ea50iver5+arth6
. 7 *on-$pen 4Close 8it5>ater (ealed5(eptic Tan,6
DDrn,>CC
1 7 8iped >ater (ource
. 7 *on 8iped >ater (ource
D#ightCF
1 7 +lectricity
D
. 7 *ot +lectricity
DHuelCG
1 7 #8/
. ' *ot #8/ 4Hirewood5Charcoal6
D>DispC2
1 7 "unicipal /arbage Truc,
. 7 *ot /arbage Truc,
D#&seE1
1 7 $ther then 0esidential &se
. 7 *ot $ther than 0esidential &se
Multile Regression Analysis
In finding out the relationship between location and quality of life, the statistical
procedure ' multiple regression analysis was used to examine the relationship
between household monthly income and several other variables li,e distance from
city heart, household si)e, number of income earners, amount of land occupied,
duration of occupation, household monthly expenditure and household transport cost
for wor,. In this regression analysis, <D independent variables including #$C!TI$*
along with other important variables were used as independent variables and monthly
household income was used as the dependent variable. The obIective of the use of this
dependence multivariate analysis technique was to predict monthly household income
of the informal settlements, as income is an important indicator of quality of life of
the informal settlements. Considering income as a crucial factor for the urban poor,
the present investigation utili)ed the techniques of multiple regression analysis in
F
order to ran, the importance of the relevant independent variables. In multiple
regression analysis, it was intended to run separate regression analysis for two
denominations of location ' danger area and non-danger area to see the effects of
location on household income.
The regression equation applied here was-
9 7
1
:
.
;
.
:
<
;
<
: = :
1
;
n
?
>here, 9 7 %ousehold "onthly Income

1


7 The Intercept

i
7 The 0egression Coefficients
;
i
7 The +xplanatory Bariables
Dependent Bariable
%%IncC3 7 %ousehold Income
Independent Bariables
#oc< 7 #ocation
. 7 Danger !rea
1 7 *on-Danger !rea
DistC 7 Distance from Central usiness District5Downtown
%%(iD 7 %ousehold (i)e
ChildF 7 *umber of Children
G
$"emG 7 *umber of $ther %ousehold "embers
!geE 7 !ge of the %ousehold head
8"ov<< 7 *umber of 8revious "oves
Hl(pc<G 7 Hloor !rea
0oom<2 7 *umber of 0ooms
Inc+CE 7 *umber of Income +arning "embers
Tra>,FE 7 "onthly %ousehold Transport +xpenditure for >or,place
+xHdD3 7 "onthly %ousehold +xpenditure for Hood
+xClF1 ' "onthly %ousehold +xpenditure for Clothing
+x+dF. 7 "onthly %ousehold +xpenditure for +dication
%0entF< 7 "onthly house 0ent
+x"dFC 7 "onthly %ousehold "edical +xpenditure
+x#iFD 7 "onthly +xpenditure for #ighting
+x>aFF 7 "onthly +xpenditure for >ater
+xHuFG 7 "onthly +xpenditure for Huel
+x0eF2 7 "onthly %ousehold +xpenditure for 0ecreation
#0entF3 7 "onthly 0ent for #and
%%+xG1b 7 !verage "onthly %ousehold +xpenditure
#!mt2G 7 !mount of #and $ccupied
Du$c22 7 Duration of $ccupation
LOGISTIC REGRESSION ANALYSIS
The logistic regression showed that location of informal settlements in danger area
and non-danger area can be explained by six of the .E variables considered in the
2
stepwise selection procedure in the (!( 4(tatistical !pplication (ystem6 system. The
Table G.. presents the results of the parametric estimation for the logistic regression
mode. 0esults of the logistic regression analysis showed 3G.E percent concordant
4Table6. (ix significant factors affect location of informal settlements in danger area
or non-danger area at .1J level of significance.
Table- G..- !nalysis of "aximum #i,elihood +stimates
!aria"le Para#eter
Esti#ate
Stan$ar$
Error
%al$
C&is'uare
Pro"a"ility
Intercept D.DC<< ...31E .C.EFC3 1.111<
Distance
from CD
- 1.<13D 1.122< 2.CGCE 1.11G2
Transport
Cost for
>or,
- 1.11.CD 1.111FC3 G..FCF 1.1.C.
#and
$wnership
- D.122D ..G3<E F.E1.G 1.1.G1
%ouse Type - ..3221 1.ECF2 F.F3G3 1.1.E1
(anitation
Type
- C.DEG3 ..13D1 .1..FE3 1.11.D
#and &se <.F<FE 1.ECDG 3..FE. 1.11<F
Bariables with negative co-efficient are DI(T!*C+ H0$" CD 41.<13D6,
T0!*(8$0T C$(T H$0 >$0K 41.11.CD6, #!*D $>*+0(%I8 4D.122D6,
%$&(+ T98+ 4..32216, and (!*IT!TI$* T98+ 4C.DEG36, and with positive co-
efficient is #!*D &(+ 4<.F<FE6.
Table G.<- !ssociation of 8redicted 8robabilities and $bserved 0esponses
Items Balues
8ercent Concordant 3C.E
8ercent Discordant G..
Tied 4<<D1 pairs6 1..
E
(omersL D 1.EF1
/amma 1.E2E
Tau-a 1.CE.
c 1.3CE
The co-efficient of distance from CD implies that for each unit increase in the
distance of informal settlements from central business district the log odds of being in
danger area increases by 1.<13D. #i,ewise in case of transport cost for wor,, each unit
decrease in transport cost for wor, increases the occurrence of being location in
danger area. Hor land ownership, government owned land increases the occurrence of
being in danger area. The occurrence of being in danger area increases the concrete
house type. !nd for sanitation type, the occurrence of location in danger area
increases the occurrence of open type sanitation 4sea5river5open pit6. The coefficient
of land use implies that the occurrence of location being in danger area increases the
land use only for residential purposes.
Distance of informal settlements from central business district 4CD6, household
monthly transport cost for wor,, land ownership, house type, sanitation type and land
use are the significant variables explaining location of informal settlements in
Tacloban City. Hrom the research it is clear that the values on probability of location
being in danger area is higher for low distance of settlements from CD, low
transportation cost for wor,, government owned land and concrete type of house and
open type of sanitation. The most li,ely probabilities range from 1.33DF. to 1.33E3F
4Table G.C6. The least li,ely probabilities ranging from 1.1111. to 1.111.F in terms of
location of being in danger area. The most li,ely and least li,ely probability of being
3
non-danger area is Iust the reverse of the least li,ely and most li,ely of the probability
of being in danger area. Thus in terms of both location of being danger and non-
danger area, the most li,ely probabilities of being in non-danger area for six variables
are the most li,ely and the least li,ely probabilities of being in danger area the least
li,ely probabilities.
The computed probabilities for being in danger area explains that informal settlements
with the distance of half 41.F6 ,ilometers from CD, no household transport
expenditure, government-owned land, house type built of strong materials, open
sanitation type and not other than residential land use has the highest probability
41.33E3F6 to be in danger area.
Table G.C- Computed 8robabilities for "ost #i,ely and #east #i,ely of #ocation in
Danger !rea
DISTANCE
FROM
CBD
TRANSPORT
COST FOR
WORK
LAND
OWNERSHIP
HOUSE
TYPE
SANITATION
TYPE
LAND
USE
PROBABILITY
Most Likely
0.5 0 0 0 0 1 0.99895
1 0 0 0 0 1 0.99883
.1
2 0 0 0 0 1 0.99856
0.5 500 0 0 0 1 0.9979
1 500 0 0 0 1 0.99771
2 500 0 0 0 1 0.99718
0.5 1000 0 0 0 1 0.99598
1 1000 0 0 0 1 0.9955
8 0 0 0 0 1 0.9995
2 1000 0 0 0 1 0.9951
Least Likely
8 3000 1 0 1 0 0.00015
13 1000 1 1 1 0 0.00010
0.5 3000 1 1 1 0 0.00010
1 3000 1 1 1 0 0.00009
8 2000 1 1 1 0 0.00008
2 3000 1 1 1 0 0.00007
13 3000 1 0 1 0 0.00005
13 2000 1 1 1 0 0.00003
8 3000 1 1 1 0 0.00002
13 3000 1 1 1 0 0.00001
In computation of probabilities for the logistic regression model, variables are
considered in following ways.
..
Hor #$C!TI$*
. 7 Danger !rea
< 7 *on-Danger !rea
Hor DI(T!*C+ H0$" CD
4K"6
1F ,m
. ,m
< ,m
E ,m
.C ,m
Hor T0!*(8$0T!TI$* C$(T
H$0 >$0K 48eso6
1 8eso
F11 8eso
.111 8eso
<111 8eso
C111 8eso
Hor #!*D $>*+0(%I8
. 7 *on /overnment
< 7 /overnment
Hor %$&(+ T98+
. 7 >ith Concrete uilding "aterials
1 ' >ithout Concrete uilding "aterials
Hor (!*IT!TI$* T98+
. 7 $pen Type 4$pen 8it5(ea50iver5$pen
(pace6
< 7 Close Type 4Close Type5>ater
(ealed5Hlush6
Hor #!*D &(+,
. 7 *ot $ther than 0esidential
1 7 $ther than 0esidential
()*) M+LIPLE REGRESSION ANALYSIS

In multiple regression analysis, household income was considered as the dependent
variable provided that income is a very important indicator of quality of life. $ut of
<C independent variable including location, only six variables in stepwise selection
procedure explain household income in informal settlements. . The results of the
regression analysis are presented in Table G.D. 0esults of the regression analysis
showed that location has no significant relationship with household income. The
.<
results provided that at .1 percent significance level six regressors predict household
income in informal settlements
Table G.D- +ffects of %ousehold +xpenditure, *umber of Income +arners, %ousehold
+xpenditure for Clothing, %ousehold (i)e, %ousehold +xpenditure for Hood and
%ousehold +xpenditure for #ighting on %ousehold Income
!aria"le Regression Coe,,icients Stan$ar$
Error
-eta t P
4Constant6 -.1<2.1F< .1.F.<F< -..1.< .C.D
%ousehold +xpenditure .EDD .<.. .D<2 C.33F .111
*umber of Income
+arning "embers
..12.CFG C33.CFC ..EG <.22C .112
%ousehold +xpenditure
for Clothing
F.FGE <..G2 ..2F <.F21 .1.<
%ousehold (i)e -CE<.C.2 .FG.G1C -..G< -<.DD. .1.2
%ousehold +xpenditure
for Hood
.22. .CC< .<C1 <.C<1 .1<C
%ousehold +xpenditure
for #ighting
<.GCF ..<EE ..D. <.1D2 .1DC
Hrom the results of multiple regression analysis it was found that household income
is significantly and positively elated to household expenditure, household
expenditure number of income earners in the household, household
expenditure for clothing and household expenditure for lighting and
household si)e has negatively related to household income. In order to see
whether location is a significant factor of household income, it was intended
if significant, to run separate regression for each denomination of location '
danger area and non-danger area, but the result showed location has no
significant effects on household income.
.C
The t ' statistic associated with six variables indicates that the relationship with
household income is significant and their beta coefficients indicate how strongly
related to household income, i.e. the weight or importance of the variables in
explaining the variation in the dependent variable.
The results showed that household expenditure is the most important variable
associated with household income 4beta 7 .D<26? household expenditure for food is the
second most important variable 4beta 7 .<C16 and number of income earning members
in the household is the third most important variable 4beta 7 ..EG6. The other
variables in the model are considerably less important.
().) INFL+ENTIAL FACTORS AFFECTING LOCATION OF INFORMAL
SETTLEMENTS AN/ T0EIR IMPLICATIONS
The findings of regression analyses indicate that several significant indicators or
factors of quality of life can be explained by location of informal settlements in
danger area and non-danger area. It was hypothesi)ed that factors affecting location
and effects of location on quality of life are not same for informal settlements located
in danger area and non-danger area. The logistic regression analysis and multiple
regression analysis of this study have examined the hypotheses regarding the
relationship between location and quality of life indicators of informal settlements.
(ix of the .E variables considered in the logistic regression analysis significantly
determine location of informal settlements. Informal settlement households with high
probabilities of being located in danger area are those within short distance from
CD, lower transportation expenditure, government owned land occupancy, strong
.D
materials built house type, open sanitation type and only residential use of land.
Informal settler households with lower probabilities of being in danger area are those
who are located far from the CD, spend more transport expenditure for wor,,
occupies privately owned land, construct not concrete house type, use close sanitation
type and use land for other than residential purposes. The highest probability of being
in danger area 41.33E3F6 is explained by 1.F ,ilometer distance of informal
settlements from CD, 1 8h8 travel cost for wor,, government owned land
occupancy, strong materials built house type, open sanitation type and only residential
use of land. The lowest probability of being in danger area 41.1111.6 can be explained
by .C ,ilometer distance from CD, C111 8h8 travel cost for wor,, privately owned
land, light materials built, close sanitation type, and other than residential use of land.
The findings of the multiple regression analysis indicate that household expenditure,
number of household income earning members, household expenditure for food,
clothing and lighting and household si)e has effects on household income in informal
settlements, although do not support the position that location has impact on
household income.
Informal settlers belong to the urban poor. #ac, of tenure, poor housing, lac, of basic
services and infrastructure characteri)e these settlements. (ince this low-income
people belonging to C1-percentile income group in the context of the 8hilippines,
cannot find alternative to squatting in danger area to find shelter in the city which
provides their subsistence livelihood. (ince most of the vacant land in danger area i.e.
along shoreline, riverban,s, cree,s etc are owned by the government and within city
proper in Tacloban City.
.F
!s from rural areas in Tacloban City, #eyte and neighboring (amar 8rovince, people
come to city for greener pasture and due to inaccessibility to land and unafforability to
buy or rent lot and housing, the informal settlements develop along shorelines,
riverban,s etc. "ost of these settlers have been staying for long time and local
government already put much investment for basic services and infrastructure side by
side householdsM investment in housing. !s these areas are vulnerable to typhoon and
flood, they often li,e to build their houses with strong materials to protect from
natural disasters.
!s their main concern is minimum subsistence, they do not thin, much on many
aspects of quality of life such as sanitation, garbage disposal, environmental pollution
etc. This scenario is also true for informal settlements in general. %owever the study
shows that for informal settlements located in danger areas deserve attention for
sustainable development of this large urban population considering their contribution
to the economy. $ver time the public policies has intended to address informal
settlement problem in various ways and resettlement of informal settlements from
danger areas is an important component of the relevant policies in the context of the
8hilippines.
The findings of the study have shown location of informal settlements has various
implications to their quality of life. #ocation of informal settlements whereas
considered negatively, deserves serious and more attention than it has received. The
reason for attention to location of informal settlements can be Iustified by the study
findings on various grounds.
.G

You might also like