Science Journal of Public Health

2015; 3(5): 707-718

Published online July 31, 2015 (http://www.sciencepublishinggroup.com/j/sjph)
doi: 10.11648/j.sjph.20150305.27
ISSN: 2328-7942 (Print); ISSN: 2328-7950 (Online)

Modeling the Determinants of Time-to-age at First Marriage

in Ethiopian Women: A Comparison of Various Parametric
Shared Frailty Models
Bedasa Tessema1, Salie Ayalew2, Kasim Mohammed2, *
1
Department of Statistics, College of Natural & Computational Sciences, Drie-dawa University, Drie-dawa, Ethiopia
2
Department of Statistics, College of Natural & Computational Sciences, University of Gondar, Gondar, Ethiopia

Email address:
t_bedasa@yahoo.com (Bedasa T.), salie.ayalew@gmail.com (Salie A.), kasim.fti@gmail.com (Kasim M.)

To cite this article:

Bedasa Tessema, Salie Ayalew, Kasim Mohammed. Modeling the Determinants of Time-to-age at First Marriage in Ethiopian Women: A
Comparison of Various Parametric Shared Frailty Models. Science Journal of Public Health. Vol. 3, No. 5, 2015, pp. 707-718.
doi: 10.11648/j.sjph.20150305.27

Abstract: Marriage is an important part of human life and age at first marriage is the age at which individuals get married.
This varies across communities and individuals in different country. Ethiopia is one of the Sub-Saharan Africa in which highest
at early marriage and a small number of delayed marriages are occurred. Survival analysis is a statistical method for data
analysis where the outcome variable of interest is the time to the occurrence of an event. Frailty model is an extension of Cox's
proportional hazard model in which the hazard function depends upon an unobservable random quantity, the so-called frailty.
Regional states of the women were used as a clustering effect in all frailty models. The study aimed to model the determinants
of time-to-age at first marriage in Ethiopia. The data source for the analysis was the 2011 EDHS data collected during
September 2010 through January 2011 from which the survival information of 12208 woman on age at first marriage. The
gamma and inverse Gaussian shared frailty with exponential, Weibull and log-logistic baseline models was employed to
analyze risk factors associated with age at first marriage using socio-economic and demographic factors. All the fitted models
were compared by using AIC. Out of the total, about 69.3% of women were married and 30.7% were not married at different
age of marriage. The median of age at first marriage was 17 years. The log-logistic with inverse Gaussian shared frailty model
had minimum value of AIC when compared with other models for age at first marriage dataset. The clustering effect was
significant for modeling the determinants of time-to-age at first marriage dataset. Based on the result of log-logistic-inverse
Gaussian shared frailty model, women educational level, head/parents occupation, place of residence, educational level of
head/parents, access to media and respondent work status were found to be the most significant determinants of age at first
marriage. The estimated acceleration factor for the group of women's who had secondary and higher educational level were
highly prolonged age at first marriage by the factor of ϕ=1.0796 and ϕ=1.1497 respectively. The log-logistic with inverse
Gaussian shared frailty model described age at first marriage dataset better than other models and there was heterogeneity
between the regions on age at first marriage. Improving girls and young women access to education was an important avenue
for rising women's age at first marriage and for empowering women.
Keywords: Time-to-age at First Marriage, Risk Factors, Comparison of Models

1. Introduction
Marriage is an important family institution for the interdependent. For the society as a whole, it unites several
individual and the society at large. For the individual, it is a individuals from different families and represents the creation
significant and memorable event in one’s life cycle as well as of a production and consumption unit as well as one for the
the most important foundation in the family formation exchange of goods and services. In addition, marriage marks
process. It is also a rite of passage that marks the beginning the beginning to an end of the transition to adulthood as the
of an individual’s separation from the parental unit, even if individual separates from the parental home, even if
generations continue to be socially and economically generations continue to be socially and economically
 Science Journal of Public Health 2015; 3(5): 707-718 708

interdependent through the extended family. A marriage is a Niger, 47 percent of women aged between 20 and 24 were
legally recognized union between a man and a woman in married before the age of 15, and 87 percent before the age
which they are united sexually; cooperate economically, and of 18, a total of 53 percent had also had a child before the age
may have children through birth or adoption (Ikamari, 2005). of 18. Bayisenge (2010) observed that African women in
Age at marriage is the age at which individuals get married general marry at a much earlier age than their non-African
and this varies across communities and individuals. Marriage counterparts, leading to early pregnancies. In average, age at
a time may be wanted or unwanted at a particular time. The first marriage is relatively high, compared with developed
term “age at early marriage” is used to refer both formal countries and many other developing countries.
marriages and informal unions in which a girl lives with a Ethiopia has one of the highest rates of age at early
partner as if married before age of 18 (UNICEF, 2005; Forum marriage in Sub-Saharan Africa. A study by the National
on Marriage and the rights of women and girls, 2001). Committee on Harmful Traditional Practices of Ethiopia
According to UNFPA (2006), early marriage, also known as (NCTPE) estimated the proportion married before the age of
Child marriage, is defined as "any marriage carried out below 15 are 57 percent. The same study shows that the practice
the age of 18 years, before the girl is physically, occurs in its most extreme forms in northern Ethiopia, where
physiologically, and psychologically ready to shoulder the girls are married as young as eight or nine years of age.
responsibilities of marriage and childbearing". Age at Child Although age at early marriage is widely practiced in many
marriage, on the other hand, involves either one or both parts of the country, rates in Amhara and Tigray region are
spouses being children and may take place with or without much higher than the national average (82 percent in Amhara,
formal registration, and under civil, religious, or customary 79 percent in Tigray, 64 percent in Benshangul, 64 percent in
laws. Age at early marriage is common in much of the Gambella and 46 percent in Afar) (NCTPE, 2003). A recent
developing world, adolescent and age at child marriage study conducted in two woreda's of the Amhara region also
continues to be a strong social norm, particularly for girls. It shows that 14 percent of women were married before age of
is associated with early childbearing, in most cases 10 years, 39 percent before age of 15 years, and 56 percent
particularly in the developing world; the main purpose of before age of 18 years (Population Council, 2004).
marriage is to have children (UNFPA, 2006). Age before 18 years marriage stands in a direct conflict
According to Demographic and Health Surveys (DHS), with the objectives of the Millennium Development Goals
which provide much of the current country-level child (MDGs) (Mathur et al., 2003). It threatens the achievement
marriage data, age at child marriage is most common in the of MDGs such as eradicating extreme poverty and hunger,
world’s poorest countries. The highest rates are in sub- achieving universal primarily education, promoting gender
Saharan Africa and South Asia as well as parts of Latin equality and empowering women, reducing child mortality,
America and the Caribbean (NRC/IOM, 2005). A UNICEF improving maternal health and combating HIV/AIDs,
study found that 48 percent of women were married before malaria and other diseases (UN, 2007).
15 years and 24 were married before 18 years in South Asia. In this study, we used shared frailty models by assuming
The prevalence of age at early marriage is 42 percent in that marriage within the same cluster (region) shares similar
Africa (UNICEF, 2005) and more than 60 percent in some risk factors, which could be taken care of the frailty term at
parts of East and West Africa (IPPF and UNFPA, 2006). In regional level. This model is a conditional independence
Latin America and the Caribbean, prevalence is 29 percent, model where the frailty is common to all individuals in a
though some individual countries have much higher rates of cluster and therefore responsible for creating dependence
age at early marriage (UNICEF, 2005). Also age at child between event times. Parametric frailty models are used to
marriage is common in the Middle East, where nearly half of investigate the relationship between different potential
girls younger than 18 in Yemen and Palestine are married covariates and time-to-age at first marriage for clustered
(IPPF and UNFPA, 2006). In sub-Saharan Africa, for survival data with a random right censoring.
example, 21 of 30 countries have seen an increase in the
national age at marriage over the past several decades 1.1. Statement of the Problem
(Westoff, 2003). This increase in the age at marriage is Many scholars recommend the need to conduct in-depth
occurring slowly and unevenly within countries, however, studies on the risk factors of age at marriage among women
and many girls are missed by this trend. According to for both developing and developed countries. Age at early
(UNICEF's, 2011) figures, 66 percent of Bangladesh girls are marriage is a health issue as well as a human right violation.
married before the age of 18 and approximately a third of A recent review show that girls who marry before the age of
women were married by the age of 15 ; although the legal 18 are disproportionately affected by complicated
age at first marriage for females in Bangladesh is 18 years. pregnancies that may lead to maternal mortality and
The highest rates of child marriage are found in West morbidity: girls aged 10–14 are five times more likely to die
Africa, in countries such as Niger, Chad, and Mali. However, in pregnancy or childbirth than women aged 20– 24; girls
in East Africa, the numbers of girls married in countries such aged 15–19 are twice as likely to die (UNICEF, 2011). A
as Ethiopia, Zambia, and Tanzania is also substantial. In rural pregnancy too early in life before a girl's body is not fully
Tanzania, median age at marriage is 18.5. The Demographic mature is a major risk to both mother and baby. Also, they are
and Health Survey (DHS) for 1995 to 2003 shows that in more likely to experience complications of childbirth
709 Bedasa Tessema et al.: Modeling the Determinants of Time-to-age at First Marriage in Ethiopian Women:
A Comparison of Various Parametric Shared Frailty Models

including obstetric fistula and hemorrhaging (IWHP, 2009). studies related to marriage and others.
Mortality rates for babies born to mothers under age 20 are This study could provide information to government
almost 75 percent higher than for children born to older and other concerned bodies in setting policies and
mothers in Ethiopia. Teenage women are also twice as likely strategies.
as older women to die due to complications during pregnancy
and childbirth. Infants born from teenage mothers are more 1.3. Objective of the Study
likely to suffer from low birth weight, and are at higher risk The specific objectives of this study are to:-
of dying in its first year by 60% compared with infants of Identify significant factors or covariates that are
mothers in their twenties (Nour, 2006). Age at first marriage associated with time-to-age at first marriage for
has health implication for women and their under-five Ethiopian women.
children (Adebowale, 2012). Determine parametric baseline hazard, which is
In Ethiopia there is no studies that documented on the area appropriate in modeling the determinants of age at first
of age at first marriage by using parametric frailty models marriage.
except the studies were conducted on the early marriage by
using logistic regression. Many of the studies used logistic
regression analysis and Cox proportional hazard models to 2. Materials and Methods
estimate the effect of covariates on the age at first marriage; 2.1. Data source
which restricts attention to the events that occur within the
shortest time observed and the correct inference based on The data set in this study was obtained from Demographic
Cox's models needs identically and independently distributed and Health Survey data conducted in Ethiopia in 2011, which
samples respectively. Logistic regression does not account was the third comprehensive survey conducted as part of the
the censoring observations i.e., does not hold for time-to- worldwide Demographic and Health Surveys project. The
event data; however, survival analysis is more powerful than data provide in-depth information on marriage, fertility,
Logistic framework that takes censoring into considerations. family planning, infant, child, adult and maternal mortality,
A frailty model is a generalization of a survival regression maternal and child health, gender, nutrition, malaria,
model and it accounts for the presence of an unobserved knowledge of HIV/AIDS and other sexually transmitted
multiplicative effect on the hazard function by specifying diseases.
independence among observed data items conditional on a set
of unobserved or latent variables. However, the Cox 2.2. Sample Design
proportional hazards model has no such constraint and the The 2011 EDHS sample was selected using two stage
dependence of the event times is not accounted. The shared cluster design and census enumeration areas (EAs) were the
frailty model is used with multivariable survival data where sampling units for the first stage. The sample included 624
unobserved frailty is shared within groups of individuals, and EAs, 187 in urban areas and 437 in rural areas. Households
thus a shared frailty model may be thought of as a random comprised the second stage of sampling. A complete listing
effects model for survival data. However, different of households was carried out in each of the 624 selected
dependence structures result from different frailty EAs from September 2010 through January 2011. A
distributions (Hougaard, 2000). representative sample of 17,817 households was selected for
The study focuses on the modeling and identifying the the 2011 EDHS, of these, 16,702 were successfully
impact of demographic and socio-economic factors on age at interviewed. In the interviewed households 17,385 eligible
first marriage. The research questions are:- women were identified for individual interview; complete
What are the key socio-economic and demographic interviews were conducted for 16,515. Women whose current
predictors of age at first marriage amongst women in ages are 15-49 years are included in the survey. After a
Ethiopia? certain rearrangement, reorganization and removal of missing
Which baseline distributional assumption among the values the total number of women with complete information
exponential, Weibull and log-logistic; as well as frailty became 12,208.
distributions, the gamma and inverse Gaussian
distributions well describe the age at first marriage? 2.3. Variables in the Study

1.2. Significance of the Study 2.3.1. The Response (Dependent) Variable

The dependent (outcome) variable is the time to age at first
The result of this study provides information on marriage marriage. It is measured as the length of time from birth until
in Ethiopian women by analyzing the impact of different the age at first marriage which is measured in years. During
variables on survival of age at first marriage. the survey all women were asked a series of questions
Specifically; regarding to their marital status and whether they had ever
The results are expected to give some knowledge about lived with a man. The response to this question constitutes
the determinants or risk factors of age at first marriage the women age at first marriage and women who had not yet
in Ethiopian women. experienced the events resulting in right censoring of the data.
This study could be used as a landmark for further
 Science Journal of Public Health 2015; 3(5): 707-718 710

2.3.2. Predictor (Independent) Variables are respondents work status, religion, type of residence,
Several predictors are considered in this study to head/parents education level, women education level,
investigate the determinant factors for the timing of age at head/parents occupations, media exposure and wealth index.
first marriage. All of these variables are categorical. Those
Table 1. Operational definition and categorization of the covariate variables, EDHS, 2011.

Variables Definition and Categorization

Women education Women level of education (0= No education;1= Primary; 2= Secondary and 3=Higher)
Residence Place of residence for women(1=Rural;2=Urban)
Wealth index Household wealth index (1= Poor; 2=Medium; 3=Rich)
Religion Women's religion(1= Orthodox; 2= Muslim; 3= Protestant; 4= Others)
Head/parents education Education level of head/parents(0=No education;1= Primary;2= Secondary and 3=Higher)
Media Access to media(0= No; 1= Yes)
Head/parents occupations Occupational status of head/parents(1= Agriculturalists;2= Professional; 3= Laborers; 4= Business; 5= Others)
Respondent work status Working status of the respondent(0= Yes; 1= No)

h (t/x , u ) = h# (t)exp(β′x + u )
Regional state of the women would be considered as a
clustering effect in all frailty models.
where indicates the ) cluster and j indicates the * )
individual for the ) cluster, ℎ, (. ) is the baseline hazard
2.4. Methodology: Survival Data Analysis

2.4.1. Non-parametric Methods function, ui the random term of all the subjects in cluster ,
Suppose t1, t2,…, tn be the survival times of n Xij the vector of covariates for subject j in cluster , and β the
independent observations and ( ) ≤ ( ) ≤ --- ≤ ( ) , m≤ n be vector of regression coefficients.
the m distinct ordered marriage times. The Kaplan-Meier If the proportional hazards assumption does not hold, the
estimator of the survivorship function (or survival probability) accelerated failure time frailty model which assumes:-
h (t/x , u ) = h# exp(β′x + u ) exp(β′x + u )
at time t, S(t) = P(T ≥ t) is defined as:
−
( ) = = 1− If the number of subjects . is 1 for all groups, the
( ) ( ) univariate frailty model is obtained (Wienke, 2010);
With the convention that ( ) = 1 for t < t(1). In this
otherwise the model is called the shared frailty model
(Hougaard, 2000; Duchateau and Janssen, 2008) because all
equation, is the number of individuals who are at risk of subjects in the same cluster share the same frailty value.
marriage at time tj, and is the number of individuals who Let us assume Z= exp (ui) and assume Z has the gamma or
occurs an event at time tj. the inverse Gaussian distribution, so that the hazard function
The Cumulative hazard function of the KM estimator can depends upon this frailty that acts multiplicatively on it.

( )=− ( ) , where ( ) is KM estimator

be estimated as: Shared frailty models are very important in analyzing
multivariate or clustered survival data. Shared frailty model
assumes that all individuals in a subgroup or pair share the
2.4.2. Shared Frailty Model same frailty Zi (i=1, 2, ---, n), and because of this it is called
The frailty approach is a statistical modeling concept shared frailty model, but frailty from group to group may
which aims to account for heterogeneity, caused by differ. Shared frailty model is similar to the individual frailty
unmeasured covariates. In statistical terms, a frailty model is model except the only difference is that frailty is now shared
a random effect model for time-to-event data, where the among the . observations in the ) group.

ℎ. ( ) = /. ℎ, ( )exp (0 1 2. )
random effect (the frailty) has a multiplicative effect on the
baseline hazard function (Wienke et al., 2003). Vaupel et al.
(1979) used the frailty approach to derive the individual Baseline Survivor and Hazard Function
hazard function based on the population hazard function Let T be a random variable associated with the survival
obtained from life tables. The shared frailty approach times, t be the realization of the random variable T and f (t)
assumes that all failure times in a cluster are conditionally be the underlying probability density function of the survival
independent given the frailties. The value of the frailty term time t. The cumulative distribution function F (t), which
is constant over time and common to all individuals in the represents the probability that a subject selected at random
cluster, and thus it is responsible for creating dependence will have a survival time less than some stated value t, is
between event times in a cluster. This dependence is always given by:
positive in shared frailty models.

3( ) = 4(5 ≤ ) = 7 8(9):9, ≥ 0
Conditional on the random effect, called the frailty denoted
by ui, the survival times in cluster (1 ≤ ≤ n) are assumed to
be independent and the proportional hazard frailty model #

The survivor function, denoted by ( ), is defined to be the

assumes:-
711 Bedasa Tessema et al.: Modeling the Determinants of Time-to-age at First Marriage in Ethiopian Women:
A Comparison of Various Parametric Shared Frailty Models

probability of an individual surviving or being event-free Table 3. Baseline Weibull distribution for Survival and Hazard functions,

defined as ( ) = 4(5 > ) = > 8(9):9 . The survival

beyond time t (experiencing the event after time t). It is EDHS, 2011.

A(B) C(B) D(B) E(B)

∞
Parameter

@F GHI(−F K ) GHI(−F K ) @F F F, @ > 0

distribution function, that is ( ) = 1 − 3( ) and density
space
KL KL K
function is merely the complement of the cumulative

function is:- (iii). Baseline Log-logistic Distribution

−: ( )
8( ) = >0
The cumulative distribution function can be written in
:
closed form is particularly useful for analysis of survival data
with censoring (Bennett, 1983). The log-logistic distribution
The hazard function is a measure of the probability of is very similar in shape to the log-normal distribution, but is

log-logistic model has two parameters- F is the scale

failure during a very small interval, assuming that the more suitable for use in the analysis of survival data. The

parameter and @ is the shape parameter which is denoted by

individual has survived at the beginning of the interval. It is

log L ( @, λ). The distribution imposes the following

defined as:-
8( ) −: ?( )
ℎ( ) = =
?( ) :
functional forms on the density, survival, hazard and
cumulative hazard function:
Survival model is usually expressed in terms of hazard Table 4. Baseline Log-logistic distribution for Survival and Hazard functions,
function. The cumulative hazard function is defined as:- EDHS, 2011.

A(B C(B) D(B) E(B)

( ) = 7 ℎ(9):9
Parameter
space

λγt OL 1 P1
# λγt OL
λ T U, γ > 0
(1 + λt O)
K
1 + λt O 1 + λt O +Q R S
Under the parametric approach, the baseline hazard is λ
defined as a parametric function and the vector of its
parameters are estimated together with the regression By specifying one of the four functions f (t), S(t), h(t) or

The parameter F is reparameterized in terms of predictor

coefficients and the frailty parameter(s). H(t) specifies the other three functions of the above baselines.

parametric models, the shape parameter @ is held fixed.

(i). Baseline Exponential Distribution variables and the regression parameters. Typically for
The exponential distribution, with only one unknown
parameter and it is the simplest of all life distribution models.
2.5. Frailty Distribution
In the exponential model, the conditional probability is
constant over time. In other words, the main feature of 2.5.1. Shared Gamma Frailty Distribution
exponential distribution is that the instantaneous hazard does The gamma distribution is very-well known and has
not vary over time. Modeling the dependency of the hazard simple densities. It is the most common distribution used for
rate on covariates entails constructing a model that ensures a describing frailty. Even though gamma models have closed
non-negative hazard rate (or non-negative expected duration form expressions for survival and hazard functions, from a

Weibull model when @ = 1. The hazard function under this

time). The exponential PH model is a special case of the computational view, it fits well to frailty data and it is easy to
derive the closed form expressions for unconditional survival
model is to assume that it is constant over time. and hazard functions.
To make the model identifiable, we restrict that
Table 2. Baseline Exponential distribution for survival and hazard functions,
EDHS, 2011. expectation of the frailty equals one and variance be finite, so
A(B) C(B) D(B) E(B)
that only one parameter needs to be estimated. Thus, the
FGHI(−F ) GHI(−F ) F F F>0
Parameter space
distribution of frailty Z is the one parameter gamma
distribution. Under the restriction, the corresponding density
(ii). Baseline Weibull Distribution function and Laplace transformation of gamma distribution:-
(Y⁄Z)\Y
]^_(LWX ⁄` )
,c > 0
WX
8V(V. ) =
Weibull distribution is one of the parametric distributions
a( b` )` Y⁄Z
which are used for the analysis of life time data and mostly
used in literature for modeling life time data (Ibrahim et al.,
2001 and Yu, 2006). The Weibull distribution is more general where Γ(. ) is gamma function. It corresponds to a Gamma
and flexible than the exponential distribution and allows for distribution Gam (µ, θ) with µ fixed to 1 for identifiability
parameter model (F J : @), where F is the scale parameter
hazard rates that are non-constant but monotonic. It is a two- and its variance is θ.
and @ is the shape parameter because it determines whether
The associated Laplace transform is:-

d(?) = (1 + ?bc)L` , c > 0

the hazard rate increases when, @ > 1 and decreases when @
the hazard is increasing, decreasing, or constant over time i.e.,

< 1 as time goes on. When @ = 1, the hazard rate remains Note that if θ > 0, there is heterogeneity. So the large
constant, which is the special case of exponential. values of θ reflect a greater degree of heterogeneity among
 Science Journal of Public Health 2015; 3(5): 707-718 712

groups and a stronger association within groups. The For identifiability, we assume z has expected value equal
conditional survival function of the gamma frailty to one and variance c.
distribution is given by: (Gutierrez, 2002). The Laplace transformation of the inverse Gaussian

` ( ) = e(1 − c f ( )g)h
L b
` c >0
distribution is:-
Y
L( n `t) bs
L(s) = exp P S c > 0, ? > 0
The conditional hazard function of the gamma frailty `
distribution is given by: (Gutierrez, 2002)
ℎ` ( ) = ℎ( )e1 − θlnfS(t)ghL
For the inverse Gaussian frailty distribution the conditional
survival function is given by: (Gutierrez, 2002).
1
`( ) = GHI u o1 − e1 − 2c f ( )gh b
qw c > 0
c
where S (t) and h(t) are the survival and the hazard functions
of the baseline distributions.
Larger variance indicates a stronger association within
groups. For the Gamma distribution, the Kendall's Tau For the inverse Gaussian frailty distribution the conditional
(Hougaard, 2000), which measures the association between hazard function is given by: (Gutierrez, 2002).

ℎ` ( ) = ℎ( )e1 − 2c f ( )gh c > 0

any two event times from the same cluster in the multivariate L b
case. It is an overall measure of dependence and independent
of transformations on the time scale and the frailty model where S (t) and h(t) are the survival and the hazard functions
used. The associations within group members are measured of the baseline distributions.
by Kendall's, which is given by:- With multivariate data, an Inverse Gaussian distributed

m = (`n ) , where mT(0,1)

` frailty yields a Kendall's Tau given by:-

, where mT(0, 1/2)

xyz o b` q | xyz (L{)
m = − +2 > b :9
2.5.2. Inverse Gaussian Shared Frailty Distribution ` `s ` {
Similar to the gamma frailty model, simple closed-form
2.6. Method of Parameter Estimation
expressions exist for the unconditional survival and hazard
functions, this makes the model attractive. The probability Under the assumption of right-censoring and of
density function of an inverse Gaussian shared distributed independence between the censoring time and the survival
random variable with parameter θ > 0 is given by time of random variables, given the covariate information,
b
q, c > 0, V > 0
L(WX L )s
8V(V. ) = o q V. GHI o
Lrb
the marginal log-likelihood of the observed data can be
p` `WX

t …. | ….

}J ~(ℎ, , 0, c; /, 2) = €• (ℎ, (‚ *)GHI(H * 0))

ƒ „.
‡27 / ˆ.
GHI •−/ ‰ , (‚ *)GHI(2 * ƒ 0)‡ 8(V ):V Š
.† † # †
t …. ….

= €• (ℎ, (‚ *)GHI(2 * ƒ 0))„. ‡ 2(−1)ˆ. d(ˆ.) •‰ # (‚ *)GHI(H * ƒ 0 ‡Š

.† † †

Taking the logarithm, the marginal likelihood is

‹Œ• (ℎ, (. ), 0, c; /, 2) = ∑t.• ∑….† • *‘ ’~‘ℎ, (‚ *)“ + 2. ƒ 0“ + ’~ (−1)ˆ. d(ˆ.) ‘ ∑….† , (‚ *)GHI(H * ƒ 0) “ ”-

where : = ∑….† • * is the number of events in the ) θ are obtained by maximizing the marginal log-likelihood of
clusters and d (•) (. ) is the – ) derivative, the Laplace
order derivatives d(•) (.) of the Laplace transform up to q =
the above. This can be done if one is able to compute higher
transform of the frailty distribution Z is defined as:-

d(t) = —eGHI(−/?)h = ># GHI(/. ?)8(/. ):V. ,t˜# and

|
max {d1, ---, ds}. Symbolic differentiation is performed in R,
but is impractical here; mainly because this is very time
|
consuming Munda et al. (2012).

d (•)
(?) = (−1)• 7 / • GHI(−/?)8(V):V , – ≥ 0 2.7. Comparison of Models
#

where ℎ, (. ) represents a vector of parameters of the baseline

Model comparison and selection are among the most

hazard function, 0 the vector of regression coefficients and θ

common problems of statistical practice, with numerous

the variance of the random effect. The estimates of ℎ, (. ) , β,

procedures for choosing among a set of models (Kadane and
Lazar, 2001) and (Rao and Wu, 2001). There are several
713 Bedasa Tessema et al.: Modeling the Determinants of Time-to-age at First Marriage in Ethiopian Women:
A Comparison of Various Parametric Shared Frailty Models

methods of model selection. The most commonly used to incorporate the parametric form of the baseline hazard

The Cox-Snell residual for the * ) individual with

methods include information criteria. One of the most rates (Klein and Moeschberger, 2003).
commonly used model selection criteria is Akaike
Information Criterion (AIC). A data-driven model selection observed survival time is given by †•(ƒž ⁄Ÿž )†L ¡¢ £(ƒž ⁄Ÿž ),
method such as an adapted version of Akaike's information where and are the estimated values of the cumulative
criterion AIC (Akaike, 1974) is used to find the truncation hazard and survivor function of the * ) subject at time
point of the series. In some circumstances, it might be useful respectively. If the model fits the data, then the ′? should
to easily obtain AIC value for a series of candidate models have a standard ( F =1) exponential distribution, so that a
(Munda et al., 2012). In this study, we used the AIC criteria
cumulative hazard of the ′? should be a straight line with
hazard plot of versus the Nelson–Aalen estimator of the
to compare various candidates of parametric frailty models.
The model with the smallest AIC value is considered a better
slope unity and zero intercept. If yes, the fitted model is
fit.
adequate. In general, Cox-Snell residual that provides a
2.8. Model Diagnostics (Checking) check of the overall fits of the model (Cox and Snell, 1968).
The three baseline hazard functions of Cox–Snell residuals
2.8.1. Evaluation of the Baseline Parameters that are considered in this study are given below:
The graphical methods can be used to check if a
parametric distribution fits the observed data or not. The Table 5. The three baseline hazard functions of Cox–Snell residuals, EDHS,
model with the Weibull baseline has a property that the log (- 2011.

F K ). Hence, log (-log ( (t))) = log (F ) + @ log (t). The

log ( (t)) is linear with the log of time, where (t) = exp (-
F exp (0′2 )
Model

intercept and slope of the line will be rough estimate of log F F K¤ exp (0 ′2 )
Exponential

and @ respectively. This property allows a graphical 1

Weibull

1 + F ¤
K
exp (0′2 )
Log-logistic

plotting log (–log ( ?̂ ( )) )) versus log (t) where (t) is

evaluation of the appropriateness of a Weibull model by

Kaplan-Meier survival estimate (Dätwyler and Timon Stucki, 3. Results

2011).

baseline can graphically be evaluated by plotting –log (?̂ ( ))

The appropriateness of the model with the exponential 3.1. Summary Statistics

versus t where ?̂ ( ) is Kaplan-Meier survival estimate. This A total of 12208 women's was included in the study during
plot should be linear and goes through the origin (Klein, the data collection. From the total 8462(69.3%) were
experienced in marriage and the rest of 3746(30.7%) were
and hence, -log ( (t)) = F is linear with time.
1992). Because for exponential distribution, (t) = exp (-F t),
unmarried at different age of marriage between 8 years and
The appropriateness of the model with the log logistic 49 years. Furthermore, among 60.6% of women were

log (1 − ?̂ ( )⁄?̂ ( )) versus log (t), where ?̂ ( ) is Kaplan-

baseline can graphically be evaluated by plotting married before age of 18 years. This indicates that age at
early marriage is highest in Ethiopia. Regarding to

of the log-logistic model is linear with slope @ then the

Meier survival estimate. The log-failure odd versus log time educational attainment, about 48.4% of heads/parents were
illiterate while 34.0% of the heads/parents had attended
survival time follows a log-logistic distribution. Where the primary education and the remaining 17.6% was secondary
failure odds of log-logistic survival model can be computed and higher education. About 60.4% of the women
as: respondents had no work and 39.6% had a work. Women
F K
who were residing in rural area were 67.6% whereas women

1 − ?( )š?( ) = 1 + F
K
=F K
who residing in urban area were 32.4%. About 46.6% of the
1 household's wealth indexes were classified as poor while
1+F K 18.5% had medium income and 34.9% were rich.
The study revealed that educational attainments of
Therefore, the log-failure odds can be written as women; about 51.3% had no education while 36.2% had
’~((1 − ?( ))⁄?( )) = log(F K ) = log(λ) + γ
primary education and the remaining 12.5% had attended

which is the liner function of log( ).(Dätwyler and Timon

secondary and higher education. From the total number,
63.1% of heads/parents occupations were agriculturalist,
Stucki, 2011). 17.6% were professional, 12.4% were Business and 6.9%
were Laborers and Others. With regard to exposure to mass
2.8.2. The Cox- Snell Residuals media, 50.6% of the women respondent had no any access
The Cox-Snell residuals method can be applied to any of media and 49.4% was recorded for women who had any
parametric model and the residual plots can be used to check access of mass media. The minimum and maximum ages at
the goodness of fit of the model. For the parametric first marriage in the data were 8 years and 37 years
regression problem, analogs of the semi-parametric residual respectively. The median of age at first marriage was 17
plots can be made with a redefinition of the various residuals years. The skewness of age at first marriage was 1.372. This
 Science Journal of Public Health 2015; 3(5): 707-718 714

shows a data is skewed to the right distribution with the 25th, respectively. In this case it is highest when Weibull was used
50th and 75th percentiles of age at first marriage was 15, 17 as baseline next to log-logistic and it is smaller when
and 19 years respectively. exponential was used as baseline. The dependence within
The log-logistic-inverse Gaussian shared frailty model that clusters (region) for the exponential-inverse Gaussian shared
is 50131.9497 appears to be appropriate model compared frailty model (τ=0.014) and Weibull-inverse Gaussian shared
with other models. This indicates it is more efficient model to frailty model (τ=0.134).
describe age at first marriage dataset. Model comparisons were presented in Table 5.
Accordingly, it suggested that log-logistic-inverse Gaussian
Table 6. The value of AIC for Multivariable Parametric Shared Frailty shared frailty model was selected according to AIC. In this
Models, EDHS, 2011.
model all categorical variables were significant except some
Model
AIC
category of religion and wealth index of the household. From
Baseline Hazard function Frailty distribution Table 6 the confidence intervals of the acceleration factor for
Gamma 71260.6561
Exponential all significant categorical covariates do not include one at 5%
Inverse Gaussian 71260.7655
Gamma 52496.8809 level of significance. This shows that they were significant
Weibull factors for determining the time-to-age at first marriage for
Inverse Gaussian 52496.8240
Log-logistic
Gamma 50133.8556 Ethiopian women. However, from the variable of religion
Inverse Gaussian 50131.9497 category Muslims were not significant when using orthodox
as the reference category with (p-value=0.052, ϕ = 1.0113,
3.2. Multivariable Analysis
95%CI = (0.9998, 1.0228) and chi-sq = 3.78). Also, those
The variance of the frailty were significant for all baseline households who had middle wealth index was insignificant
hazard function with an inverse Gaussian shared frailty by using poor households as a reference category with (p-
distribution in the models whereas it was not significant in value = 0.74, chi-sq= 0.11, 95%CI = (0.9911, 1.0127) and ϕ
the gamma shared frailty distribution using the same baseline = 1.0018). The estimated coefficient of the parameters for
as inverse Gaussian models at 5% level of significance. This respondent who had no work status was -0.0116. The sign of
indicates the presence of heterogeneity and necessitates the the coefficients are negative which implies that decreasing
frailty models. The value of shared frailty distribution (θ) is logged of survival time and hence, shorter expected duration
0.029, 0.416 and 0.312 for exponential-inverse Gaussian, of age at first marriage of the women.
Weibull-inverse Gaussian and log-logistic-inverse Gaussian
Table 7. Results of the multivariable Log-logistic-Inverse Gaussian shared Frailty Model for age at first marriage dataset, EDHS, 2011.

Variable Coef S.e(coef) ¥ 95%CI LCL UCL Chi-sq P-value

(Intercept) 2.7976 0.0324 16.4052 (15.3958,17.4808) 7459.52 0.00
Heads/parents occupation
Agriculturalist Ref
Professional 0.0216 0.0070 1.0218 (1.0079,1.0360) 9.45 2.1e-13
Business 0.0433 0.0067 1.0433 (1.0306,1.0581) 42.17 8.3e-11
Laborers 0.0432 0.0115 1.0441 (1.0209,1.0679) 14.04 1.8e-04
Others 0.0524 0.0118 1.0538 (1.0297,1.0785) 19.91 8.1e-06
Women Education
No education Ref
Primary 0.0118 0.0047 1.0119 (1.0026,1.0212) 6.26 0.012
Secondary 0.0766 0.0089 1.0796 (1.0609,1.0986) 73.66 0.00
Higher 0.1395 0.0112 1.1497 (1.1247,1.1752) 155.9 0.00
Residence
Rural Ref
Urban 0.0169 0.0067 1.0170 (1.0038,1.0305) 6.45 0.011
Heads/parents Education
No education Ref
Primary 0.0344 0.0080 1.0350 (1.0189,1.0514) 18.51 1.7e-05
Secondary 0.0397 0.0047 1.0405 (1.0310,1.0501) 69.94 1.1e-16
Higher 0.0434 0.0098 1.0444 (1.0245,1.0646) 19.51 9.9e-06
Media
No Ref
Yes 0.0200 0.0046 1.0202 (1.0110,1.0294) 18.67 1.6e-05
Religion
Orthodox Ref
Muslims 0.0112 0.0058 1.0113 (0.9998,1.0228) 3.78 0.052
Protestant 0.0213 0.0071 1.0215 (1.0074,1.0358) 8.79 0.0027
Others 0.0309 0.0138 1.0314 (1.0039,1.0597) 5.03 0.025
Wealth Index
Poor Ref
Middle 0.0018 0.0055 1.0018 (0.9911,1.0127) 0.11 0.74
715 Bedasa Tessema et al.: Modeling the Determinants of Time-to-age at First Marriage in Ethiopian Women:
A Comparison of Various Parametric Shared Frailty Models

Variable Coef S.e(coef) ¥ 95%CI LCL UCL Chi-sq P-value

Rich 0.0477 0.0053 1.0489 (1.0380,1.0598) 82.85 0.00
Work status
Yes Ref
No -0.0116 0.0043 0.9885 (0.9802,0.9968) 7.25 0.0071
Log(scale) -2.1305 0.0090 0.00
Frailty 104.42 0.00
θ = 0.312 λ = 6.1669e-11
τ =0.109 γ = 8.4034 AIC= 50131.9497

Coef= coefficient, S.e= standard error, ϕ = acceleration factor, 95% CI=Confidence Interval for acceleration factor, LCL=lower class limit, UCL= upper class
limit, Chi-sq= Chi-square, Ref=Reference, θ = variance of the random effect, λ = scale parameter, γ = shape parameter, τ = Kendall's Tau.

The occupational status of heads/parents was statistically clusters (region) is measured by Kendall's tau is τ=0.109.
determine age at first marriage of the women. The time rate
and 95% Confidence interval of acceleration factors for 3.3. Survival Function of Different Categorical Group of
occupational status of heads/parents for a group of Covariates
professional, business, laborers and Others was In all frailty models the categories of women's educational
1.0218(1.0079, 1.0360), 1.0433(1.0306, 1.0581), level were highly significant at 5% level of significance
1.0441(1.0209, 1.0679) and 1.0538 (1.0297, 1.0785) when when compared with the reference category (no education).
compared to occupation of agriculturalists group (as The gap between the four curves distinguishes that the
reference category) respectively. survival distribution of age at first marriage for Ethiopian
The 95% confidence interval for acceleration factor of women. The differences that are displayed in survival curve
women educational levels was (1.0026, 1.0212), (1.0609, emphasize that women who had higher education was
1.0986) and (1.1247, 1.1752) for the group of primary, married later when compared with others and women who
secondary and higher education's respectively. This are uneducated had less survival than educated. As we
confidence interval does not include one in all; indicating indicate from the graph there is a high gap at the mid time
primary, secondary and higher education's were significantly between a primary and secondary educated woman on
important factors for the timing of age at first marriage by marriage. This shows women who had secondary education
using uneducated women as a reference category. were more survived than uneducated and primary educated
Accordingly, it prolonged the age at first marriage by a factor women.
of (ϕ =1.0119, ϕ = 1.0796 and ϕ =1.1497) for primary,
secondary and higher education respectively at 5% level of
significance.
The acceleration factor for women who are lived in urban
area was 1.0170 times greater than those who are lived in
rural area (Ref) (ϕ: 1.0170, 95%CI: 1.0038, 1.0305). The 95%
confidence interval of the acceleration factor for those
women who had any access of media and don't had works are
(1.0110, 1.0294) and (0.9802, 0.9968), its acceleration
factors are 1.0202 and 0.9885 by using those not had any
access of media and had a work as a reference category at 5%
level of significance respectively.
The coefficients on the categorical variable of
heads/parents educational level, shows that the survival time
of age at first marriage increased with changing from one
category to another (primary, secondary and higher) Figure 1. Displays the survival function for the group of women educational
educational level relative to those heads/parents with no level on age at first marriage by using log-logistic-inverse Gaussian shared
frailty model, EDHS, 2011.
education as a reference group and the survival times was
lengthened by 3.50%, 4.05% and 4.44% respectively for the
3.4. Discussion of Results
group of primary, secondary and higher educational level of
heads/parents. The findings of this study revealed that the educational
The estimate of shape parameter in the log-logistic-inverse level of women had a significant effect on the survival of age
Gaussian shared frailty model is (γ=8.4034). This value at first marriage with 5% level of significance and it
shows the shape of hazard function is unimodal because the prolonged age at first marriage by the factor of ϕ= 1.0119,
value is greater than unity i.e., it increases up to some time 1.0796 and 1.1497 for primary, secondary and higher
and then decreases. The heterogeneity in the population of education respectively when illiterate women was used as the
the region which is used as a clusters are estimated by our reference group. The result of the study shows that woman
selected model is θ=0.312 and the dependence within the who had higher education was more survived than those
 Science Journal of Public Health 2015; 3(5): 707-718 716

uneducated and primary education. A study conducted in that there is heterogeneity between regions by assuming
Ethiopian regions by Erulkar (2011) investigated the factors women within the same region share similar risk factors on
associated with marriage and the result suggested that marriage i.e., the correlation within regions cannot be
educational attainments of women had significant effect on ignored and clustering effect was important in modeling the
marriage and women who were not educated were married hazard function.
earlier than educated. In our study the adequacy of baseline distributions are
The results of this study suggested that place of residences checked by using graphs in figure (5). From the plot of
was significant predictive factor for age at first marriage in exponential, Weibull and log-logistic distributions; the plot of
Ethiopian women. This shows that women who lived in log-logistic was more straight line compared with
urban areas are more survived on age at first marriage than exponential and Weibull for age at first marriage dataset.
women who lived in rural areas. This might be rural areas These findings were consistent with Cox (1970), O'Quigley
tend to have institutional and normative structures such as the and Struthers (1982), Bennet (1983) and Cox and Oakes
kinship and extended family that promote early marriage and (1984) for baseline log-logistic.
childbearing, but women in urban areas need to develop
skills, gain resources, and achieve maturity to manage an 4. Conclusions
independent household and thus they have to delay marriage.
A study in Nigeria by Thomas (2010-2011) and Adebowale This study was based on a dataset of age at first marriage
(2012) found that women who are lived in rural area had a obtained from the central statistical agency of Ethiopia with
higher risk of first marriage than urban area i.e., hazard of an aim of modeling the determinants of time-to-age at first
women living in rural is greater than urban. marriage by using different parametric baseline with different
The results of this study suggest that work status of the shared frailty model on the marriage dataset of Ethiopian
women had a significant effect on age at first marriage and women. Out of the total 12208, about 69.3% were
age at early marriage was higher for women who had no experienced an event (married) and 30.7% were not
work status. This is consistent with Shapiro (1996), Zahangir experienced an event (unmarried) for a different age between
et al. (2008) and Kamal (2011), they revealed that work 8 and 49 years.
status of women were significant effect on age at first To model the determinants of time-to-age at first marriage,
marriage and pre-marital work status of women were various parametric shared frailty models by using different
significantly delayed the timing of marriage. baseline distributions were applied. Among this using AIC,
Households who had higher wealth index were found to be the log-logistic-inverse Gaussian shared frailty model is
one of the statistically significant factors from the category better fitted to marriage dataset than other parametric shared
for determining age at first marriage in our study. It showed frailty models. There is a frailty (clustering) effect on the
that age at first marriage for women was prolonged by the time-to-age at first marriage dataset that arises due to
factor of ϕ =1.0489 when we used the wealth index of poor differences in distribution of time to age at first marriage
household as (reference). This finding is consistent with among region of Ethiopia.
Kamal (2011) in Tribal women in Bangladesh and the study The result of log-logistic-inverse Gaussian shared frailty
revealed that the higher economic status of parents, the lower model showed that the factors that determine the timing of
is the probability of age at early marriage. age at first marriage are women educational level,
The result of this study also revealed that heads/parents heads/parents occupation, place of residence, educational
occupations are the important factor for age at first marriage level of heads/parents, access to any media and work status
of Ethiopian women. A similar study in Western Uganda by of the respondents are statistically significant. Also from the
Peninah et al. (2011) and Zahangir et al. (2008) in rural category of religion, protestant and others and from
Bangladesh found that the occupation of the parents were household's wealth index, the richest households are
strong socio-economic determinants of age at first marriage. significant. As educational level of the women increases, age
Also, another study in Bangladesh by Mosammat et al. (2013) at first marriage is highly prolonged in Ethiopian women.
showed that the occupations of the parents were important This indicates education of women were significant factor to
factors for determining age at first marriage. determine timing of age at first marriage and implies that
Access to mass media was found to have a significant girls should be kept in school for a longer period, not only for
effect on age at first marriage. The findings of this study the purpose of raising age at marriage, but also for biological,
showed that women who had no access of media were physical and mental maturity.
married at earlier age than those who had access of media. Awareness has to be given for the society on age at the
This finding is consistent with Tezera (2013), Zahangir and marriage. The mass media can play an effective role in this
Kamal (2011), Zahangir et al. (2008) and Joseph et al. (2012). regard and the awareness need to follow the ordinance of the
In this study, we used the region as a clustering (frailty) legal age of marriage because it is the most determinants of
effect on modeling the determinants of time-to-age at first health for women and child borne. Similarly, it is advisable to
marriage in Ethiopian women using 2011 EDHS data. The target young women, particularly those with no or little
clustering effect were significant (p-value :< 0.000) in log- education including primary school girls, with information on
logistic-inverse Gaussian shared frailty model. This showed reproductive health and to provide them to avoid ultimately
717 Bedasa Tessema et al.: Modeling the Determinants of Time-to-age at First Marriage in Ethiopian Women:
A Comparison of Various Parametric Shared Frailty Models

age early marriage. Further studies should be conducted in [16] Ibrahim J.G., Chen M. and Sinha D. (2001). Bayesian survival
each region of Ethiopia and identify other factors that are not analysis. Springer Verlag, New York.
identified in this study. Based on that study regional [17] IPPF and UNFPA. (2006). Ending Child Marriage: A Guide
government should takes an action on age at first marriage. for Global Policy Action, International Planned Parenthood
Federation (IPPF), London

Acknowledgement [18] Joseph N., M.Fajar R. and Mayang R. (2012). Prevalence of

Child Marriage and Its Determinants among Young women in
The authors of this article would like to the thank Central Indonesia, Child poverty and social protection Conference.
Statistical Agency of Ethiopia for making available the data Journal of Statistics 14, 19 - 25.
used in this research. We also thank university of Gondar for [19] Kadane J. and Lazar N. (2001). Methods and Criteria for
the use of their computers and Internate services. Model Selection. Technical Report, 759, Carnegie Mellon
University.
[20] Kamal S.M.Mostafa (2011). Socio-Economic Determinants of
Reference Age at First Marriage of the Ethnic Tribal Women in
Bangladesh, Asian Population studies.
[1] Adebowel A., Fagbamigbe A., Okareh O. and Lawal O. (2012).
Survival Analysis of Timing of First Marriage among Women [21] Klein J. (1992). Survival analysis: techniques for censored and
of Reproductive age in Nigeria: African Journal of truncated data. Medical College of Wisconsin.
Reproductive Health.
[22] Klein P. and Moeschberger L. (2003). Survival Analysis
[2] Akaike H. (1974). A new look at the statistical model Techniques for Censored and Truncated Data - Google
identification. IEEE Trans Automatic Control. Books.htm

[3] Bayisenge J. (2010). Early Marriage as a Barrier to Girl’s [23] Mosammat Zamilun Nahar, Mohammad Salim Zahangir and
Education Rwanda: National University of Rwanda. S.M. Shafiqul Islam (2013). Age at first marriage and its
relation to fertility in Bangladesh, Chinese Journal of
[4] Bennett S. (1983). Analysis of survival data by the Population Resources and Environment, 11:3, 227-235,
proportional odds model. Statistics in Medicine, 2, 273. DOI:10.1080/10042857.2013.835539
[5] Bedassa B. (2015). Risky Sexual Behavior and [24] Munda M., Rotolo F. and Legrand C. (2012). Parametric
PredisposingFactors to HIV/STI Among Students in Mizan- Frailty Models in R. Journal of American Statistical
Tepi University(Acase of Tepi Campus). Science Journal of Association: 55:1-21.
Public Health, Vol.3,No.5,Pp.605-611.
[25] Nour N. M. (2006). Health Consequences of Child Marriage
[6] Cox D. (1972). Regression models and life tables (with in Africa. Emergence Infection Disease 12(11):1644-9.
discussions). Journal of the Royal Statistical Society. 34: 187-
220. [26] Peninah A., Leonard K. Atuhaire. and Gideon Rutaremwa
(2011). Determinants of age at first marriage among women in
[7] Cox D. R. and Oakes D. (1984). Analysis of Survival Data. western Uganda. Presented in European Population
Chapman and Hall, London. Conference 2010, Vienna 1-4 September 2010

[8] Cox D. R. and Snell E. J. (1968). A general definition of [27] Population Council (2004). Supporting Married Girls: Calling
residuals with discussion. Journal of the Royal Statistical Attention to a Neglected Group.
Society. Series B 30 (1968), 248-275.
[28] Rao C. and Wu Y. (2001). On Model Selection (with
[9] Duchateau L. and Janssen P. (2008). The Frailty Model. discussion). In P. Lahiri (Ed.), Model Selection, Volume 38 of
Springer-Verlag, New York. IMS Lecture Notes - Monograph Series. Institute of
Mathematical Statistics.
[10] Erulkar A. (2013). Early Marriage, Marital Relations and
Intimate partner violence in Ethiopia. [29] Shapiro, D. (1996). Fertility Decline in Kinshasa, Population
Studies Vol50, No.1, pp 89- 103.
[11] Goncalves L., Duarte H. and Cabral M. (2015). Prevalence of
Hemoglobin S in Blood Donors in the Hospital Dr. Agostinho [30] Tezera A. (2013). Determinants of Early Marriage among
Neto, Praia City-Cape Verde. Science Journal of Public Women in Ethiopia.
Health,3:5, 600-604.
[31] Thomas E. (2010-2011). Multilevel Survival Analysis of the
[12] Hougaard P. (1984). Life table methods for heterogeneous Determinants of Age at First marriage Among Women living
populations. Biometrika 71,75-83. in Nigeria. Thesis for Master of Statistical Data Analysis, Gent
University.
[13] Hougaard P. (1986). Survival models for heterogeneous
populations derived from stable distributions. Biometrika 73, [32] UN . (2000). World Urbanisation Prospects. New York, United
387 - 396 Nations.

[14] Hougaard P. (1995). Frailty Models for Survival Data. [33] UN. (1990).Patterns of First Marriage: Timing and Prevalence,
New yorkeVol.4. No. 3 (April), pp: 221-235.
Lifetime Data Analysis. 1: 255-273.
[34] UN. (2007). The Millennium Development Goals, Report
[15] Hougaard P. (2000). Analysis of Multivariate Survival Data,
2007, United Nations, New York. Retrieved 12, August, 2013
Springer-Verlag, Newyork.
http://www.un.org/mi1lenniumgoals/pdf/mdg2007.pdf
 Science Journal of Public Health 2015; 3(5): 707-718 718

[35] UNFPA. (2003). State of World Population 2003: Making 1 frailty model with a cure fraction for modeling familial
Billion Count: Investing in Adolescents’ Health and Rights. correlations in diseases. Biometrics, 59, 1178-83.
New York
[42] Wienke A., Ripatti S., Palmgren J., Yashin A.I. (2010). A
[36] UNFPA. (2006). In ending child marriage, a guide for global bivariate survival model with compound Poisson frailty.
policy action International Planned Parenthood Federation and Statistics in Medicine 29, 275–83.
the Forum on Marriage and the Rights of Women and Girls.
U.K. [43] Yu B. (2006). Estimation of shared gamma frailty models by a
modified EM algorithm. Computational Statistics and Data
[37] UNICEF. (2005). Early Marriage: A Harmful Traditional Analysis 50, 463-474.
Practice: A Statistical Exploration. UNICEF: New York, NY.
[44] Zahangir M. S., M. A. Karim, M. R. Zaman, M. I. Hussain and
[38] UNICEF. (2011). Child Protection from Violence, M. S. Hossain, 2008. Determinants of age at first marriage of
Exploitation and Abuse. rural women in Bangladesh: A cohort analysis Trends Applied
Science. Res. 4(3):335-343, Academic journal, Department of
[39] Vaupel J., Manton K. and Stallard E. (1979). The Impact of Statistics.
Heterogeneity in Individual Frailty on the Dynamics of
Mortality. Demography. 16: 439-454. [45] Zahangir M.S. and Kamal M.M. (2011). Several Attributes
Linked with Child Marriage of Females' in Bangladesh,
[40] Westoff C. F. (2003). Trends in Marriage and Early International Journal of Statistics and Systems, volume 6.
Childbearing in Developing Countries. DHS Comparative
Reports No.5. Macro International Inc.: Calverton, Maryland.
[41] Wienke A., Lichtenstein P. and Yashin A.I. (2003). A bivariate