The Journal of Nutrition

Community and International Nutrition

Protein and Micronutrient Intakes Are

Associated with Child Growth and Morbidity
from Infancy to Adulthood in the Philippines1,2

Downloaded from https://academic.oup.com/jn/article/146/1/133/4585684 by PYEONGTAEK UNIVERSITY user on 03 October 2020

Alok Bhargava*

University of Maryland School of Public Policy, College Park, MD

Abstract
Background: Interactions between nutrition and infections in developing countries are complex, and analyses of the
interrelations require data from longitudinal studies that span several years and a comprehensive framework facilitating
the formulation of food and health policies.
Objective: The effects of dietary intakes in the Philippines for birth outcomes, heights, weights, and morbidity during ages
2–24 mo; heights and weights during ages 8–19 y; and completed adult heights were analyzed.
Methods: Data on >3000 children from the Cebu Longitudinal Health and Nutrition Survey were modeled in 4 sets of
analyses. Cross-sectional and multilevel models were estimated for length and weight at birth. Dynamic random-effects
models were estimated for childrenÕs heights, weights, and morbidity levels during ages 2–24 mo. ChildrenÕs heights and
weights during ages 8–19 y were modeled by using random-effects models. ChildrenÕs heights at age 22 y were modeled
for investigating the effects of nutritional intakes in early childhood and in adolescence.
Results: Maternal anthropometric indicators, energy intakes, and sociodemographic variables were significantly
associated with childrenÕs length and weight at birth. Dynamic models for childrenÕs heights and weights during ages
2–24 mo showed significant effects of calcium and protein intakes; child morbidity levels were significantly associated
with height and weight. Higher b-carotene intakes were significantly associated with lower morbidity levels. Analyses of
data on children during ages 8–19 y showed significant effects of protein and calcium intakes and of morbidity levels on
heights and weights. Models for childrenÕs heights at age 22 y indicated significant effects of protein and calcium intakes
during early childhood and in adolescence.
Conclusions: The results underscored the need for formulating long-term food and health policies for the Philippines that
enhance childrenÕs physical development and ultimately their adult stature, which is important for physical work capacity
and labor productivity. J Nutr 2016;146:133–41.

Keywords: anthropometric indicators, diet quality, econometric modeling, food policies, longitudinal data,
long-term evaluation, maximum likelihood estimation, morbidity

Introduction
The importance of interactions between infections and individ-
ualsÕ nutritional status has long been recognized, especially for
developing countries (1–6). Inadequate protein and micro-
nutrient intakes lower the capacity of the body to fight infections,
which, in turn, adversely affects nutritional status by prolonging
sickness spells. Moreover, absorption rates for nutrients such as
iron and calcium are often low due to excessive phytate intake
from grains (7, 8). In addition, food contamination due to poor
hygiene exacerbates nutrient loss (9). It is therefore important
1
Supported in part by the Harvest Plus program of the International Food Policy
Research Institute.
2
Author disclosures: A Bhargava, no conflicts of interest.
*To whom correspondence should be addressed. E-mail: bhargava@umd.edu.
ã 2016 American Society for Nutrition.
Manuscript received August 20, 2015. Initial review completed October 1, 2015. Revision accepted November 2, 2015.
First published online December 9, 2015; doi:10.3945/jn.115.222869.
to analyze the factors underlying childrenÕs physical growth and
morbidity that ultimately determine their adult physical work
capacity (10) and labor productivity (11).
For the design of efficacious food policies, it is important to
adopt multidisciplinary approaches that incorporate interactions
between nutritional status and infections. For example, weaning
programs for breastfed infants may not be effective if they
inadvertently increase morbidity levels because the mothers failed
to maintain hygiene in preparing the supplements (12). Moreover,
it is important for food policies to match childrenÕs nutritional
requirements with their intakes at various ages. This is an intricate
task because childrenÕs requirements depend on their ages and
growth pattern history (13, 14). It is therefore important to adopt
a long-term framework for analyzing the effects of childrenÕs
nutritional intakes and morbidity on their physical development.
133
 Furthermore, there has been emphasis in the literature on early
nutritional interventions, especially during the first 2 y of life (15).
Although such interventions may be necessary for undernourished
children growing up in poverty, it is important to adopt a broader
perspective on ‘‘critical windows’’ that may be available for catch-
up growth (16). Thus, for example, heights in the Netherlands
surpassed all countries after the Second World War presumably
due to the school milk programs, which rewarded children for
daily drinking up to 5 glasses of milk to become ‘‘M-brigadiers’’
(17). The programs were gradually discontinued in the 1980s,
and some reversion to the mean in Dutch heights is apparent
(18). Because childrenÕs requirements for protein and micronu-
trients such as calcium and iron are also high during adolescence
(3), an exclusive focus on early interventions can lead to a
misallocation of scarce resources.
For the analyses of longitudinal data on children that span
several years, it is important to tackle 2 sets of issues. First, growth
spurts, especially around puberty, are affected by the secretion of
hormones that depend on childrenÕs nutritional status, including
dietary intakes such as that of protein (19, 20). Moreover, there is
considerable variation in growth spurts during adolescence and in
the dynamics of body composition, such as leg growth, which
peaks a few months earlier than trunk growth (14). Thus, in
addition to factors such as length and weight at birth, it is useful
to account for childrenÕs dietary intakes during adolescence in
models for their completed heights.
Second, inferences from data analyses critically depend on
the models postulated for the interrelations between childrenÕs
diets, morbidity, and growth. Previous research has computed
simple correlations or associations between childrenÕs length and
weight at birth and completed heights by using data from several
countries (21). Such formulations largely ignore the nutritional
factors affecting child growth during the in-between years; the
estimated coefficients are likely to be ‘‘upward biased’’ due to the
omission of positively correlated explanatory variables.
This article analyzed the data from 21 survey rounds of the
Cebu Longitudinal Health and Nutrition Survey conducted from
1983 to 2005 involving >3000 Filipino children from birth to
adulthood (22). First, models were estimated for childrenÕs
lengths and weights at birth that were explained by socioeco-
nomic variables and mothersÕ nutritional status. Second, chil-
dren heights and weights were modeled in the period 2–24 mo of
age by estimating dynamic random-effects models (23) and the
effects of nutrient intakes were investigated; protein intakes are
important for growth, especially where the diets are lacking in
animal products (2), and calcium intakes are essential for bone
growth (24). ChildrenÕs morbidity levels in the previous 24 h and
7 d and diarrheal morbidity were modeled by using indexes that
reflected the intensity and duration of sicknesses (25, 26);
b-carotene intakes were included as explanatory variables (4,
27). Third, random-effects models were estimated by using data
at ages 8, 10, 15, and 19 y on childrenÕs heights and weights.
Fourth, childrenÕs heights when they turned 22 y were explained
by using early anthropometric measurements and protein and
calcium intakes during the ages of 2–24 mo and from 10–19 y.
Methods
The Cebu Longitudinal Health and Nutrition Survey was conducted in
metropolitan Cebu, the Philippines, from 1983 to 2005 (22); data on
maternal and infant health indicators and sociodemographic variables
are available from 21 survey rounds. At baseline, a clustering-sampling
procedure was used to randomly select 17 urban and 16 rural ‘‘barangays’’
representing ;28,000 households. Pregnant women giving birth between
134 Bhargava
May 1983 and April 1984 were invited to enroll in the study, and the refusal
rate was 4%. The 3080 women in the sample with single births were
interviewed during pregnancy, at birth, and then bimonthly from ages 2
to 24 mo (i.e., there were 13 observations on infants in the initial surveys).
Subsequently, mothers and children were interviewed in follow-up surveys
in 1991, 1994, 1998, 2002, and 2005 when the children were 8, 10, 15,
19, and 22 y old, respectively.
The first survey round conducted during pregnancy measured
mothersÕ heights and weights around the 30th week of gestation.
MothersÕ food intakes were measured by using the 24-h recall method
and were converted into energy and nutrient intakes with the use of
tables for the Philippines (28). MothersÕ food intakes were again
measured when the infants were born and when they were ;6 and 12 mo
Downloaded from https://academic.oup.com/jn/article/146/1/133/4585684 by PYEONGTAEK UNIVERSITY user on 03 October 2020
old. ChildrenÕs intakes from weaning were assessed by using 24-h recalls
from ages 4 to 24 mo. Infants were weighed at birth by using Salter dial-
faced scales, and birth length was measured with the use of locally made
boards. Infant lengths and weights and mothersÕ weights were subse-
quently measured at 2-mo intervals in 12 time periods.
Demographic and socioeconomic variables were compiled for the
households. For example, the childÕs gender and birth order and the
number of persons in the household were recorded. The highest grade
completed by the mother was also recorded. The socioeconomic status of
the household was assessed via several questions during the first 2 y.
Because of missing observations on some items, a socioeconomic status
index was constructed on the basis of 7 items: whether the family owned
the house and/or owned the land, types of vehicles owned, number of
appliances owned, number of rooms, number of furniture items owned,
and if the family owned other houses.
Furthermore, information was compiled on mother and infant
morbidities and immunizations in the surveys when children were
between 2 and 24 mo old. For mothers, the morbidity index was based
on 6 items for the past 24 h—namely, if mothers had been ill or had
diarrhea, cold or congestion, cough, fever, and other illnesses; this index
ranged from 0 to 6. For children, morbidity index 1 covered 6 symptoms
in the past 24 h: cough, fever, nasal discharge, ear discharge, sore throat,
and other illnesses. ChildrenÕs morbidity index 2 covered the past week
and inquired about 7 items: cough, nasal congestion, ear discharge, fever,
number of days of diarrhea, other illnesses, and measles; this index
ranged from 0 to 13. The childrenÕs diarrhea index was based on the
number of days the child had diarrhea and was doubled if there was
blood in the stool; this index ranged from 0 to 24. Last, childrenÕs
immunizations against diphtheria, pertussis, and tetanus, tuberculosis,
measles, polio, cholera, and other illnesses were recorded at different
time points. The total number of vaccinations received before the age of
2 y was used as an index for immunizations.
Finally, follow-up surveys were conducted when children were 8, 10,
15, 19, and 22 y old. MothersÕ and childrenÕs heights and weights were
measured in the survey rounds. ChildrenÕs food intakes were measured at
ages 10, 15, and 19 y by using 24-h recalls and were converted into
energy and nutrient intakes. ChildrenÕs total morbidities and number of
immunizations they received during the ages of 2–24 mo were included
in the longitudinal database for analyzing growth patterns when the
children were 8–19 y old.
Empirical models. The availability of longitudinal data from Cebu
enabled 4 sets of interdependent analyses. First, the cross-sectional
model for childrenÕs birth length (and weight) is given in Equation 1:
lnðLengthÞi ¼ a0 þ a1 ðGenderÞi þ a2 ðBirth orderÞI
2
þa3 ðBirth orderÞi þ a4 ðNumber of persons in householdÞi
þa5 ðNumber of persons in householdÞi
2
þa6 ðSocioeconomic indexÞi þ a7 ðMothers highest gradeÞi
þa8 lnðMothers heightÞi þ a9 lnðMothers weightÞi
þa10 lnðMothers energy intake=dÞi þ ui ði ¼ 1; 2; .; NÞ ð1Þ
The model in Equation 1 postulates nonlinearities between childÕs birth
order and number of persons in the household and the natural logarithm
of length (or weight) at birth. Moreover, mothersÕ heights and weights
are likely to be positively associated with birth size, so that it would be
inappropriate to combine them as the BMI in this model (26, 29, 30). In
addition, mothersÕ energy and nutrient intakes are important for
intrauterine development (31). The correlation between mothersÕ highest
grade and socioeconomic index was 0.41, and it did not present any
difficulties for the estimation of model parameters.
Second, the dynamic random-effects model for childrenÕs heights
with the use of bimonthly observations in the period from 2 to 24 mo is
given in Equation 2:
lnðHeightÞit ¼ b0 þ b1 ðGenderÞi þ b2 ðBirth orderÞi
þb3 ðNumber of persons in householdÞi
þb4 ðSocioeconomic indexÞi þ b5 ðMothers highest gradeÞi
þb6 ðMothers morbidity indexÞit
þb7 lnðMothers energy intake=dÞit
þb8 lnðMothers BMIÞit
þb9 ðChildren immunizations totalÞi
þb10 ðChildren morbidity indexÞit-1
þb11 ðChildren diarrhea indexÞit-1
þb12 lnðChildren protein=energy intake=dÞit-1
þb13 lnðChildren calcium=energy intake=dÞit-1
þb14 lnðHeightÞit-1
þu1it ði ¼ 1; 2; .; N; t ¼ 2; .; 11Þ ð2Þ
u1i t represents random-error terms that can be decomposed in a random-
effects fashion as follows:
u1it ¼ di þ v1it ð3Þ
where di represents child-specific random effects that were assumed to
be distributed with zero mean and constant variance, and v1it were
distributed with zero mean and constant variance (32). The estima-
tion techniques treated previous observations on childrenÕs heights as
‘‘endogenous’’ variables, i.e., correlated with the errors u1it (23) (see
below).
The dynamic model in Equation 2 contained previous measurement
on height as an explanatory variable, thereby enabling a distinction
between short- and long-run effects of explanatory variables. Moreover,
previous levels of morbidities and protein and calcium intakes were
postulated to affect current heights. Because the data were available at
2-mo intervals, the model in Equation 2 expressed nutrient intakes as
ratios to energy intakes (33, 34). Such ratios reflect diet quality and
obviate the need for separately including childrenÕs energy intakes, which
are influenced by anthropometric measures and energy expenditures
(33). It was appropriate to combine mothersÕ heights and weights as the
BMI with the use of a statistical test (26).
The models for childrenÕs weights and morbidity levels during the
ages of 2–24 mo were similar to the model in Equation 2, although there
were some differences. For example, current morbidity levels were
introduced as explanatory variables in the model for weight. This can
induce correlations between the errors affecting the model for weight
and morbidity levels and is discussed below. The models for childrenÕs
morbidity levels included explanatory variables such as householdÕs
access to piped water and flush toilets, hygiene condition, and childrenÕs
intakes of b-carotene expressed as ratios to energy intakes. Third, the
dynamic random-effects model for childrenÕs heights (and weights) with
the use of data at ages 8, 10, 15, and 19 y is given in Equation 4:
lnðHeightÞit ¼ c0 þ c1 ðGenderÞi þ c2 ðSocioeconomic indexÞi
þc3 ðMothers highest gradeÞi þ c4 lnðMothers heightÞi
þc5 lnðMothers weightÞit
þc6 ðChildren immunizations totalÞi
þc7 ðChildren morbidity index totalÞi
þc8 lnðChildren protein intake=dÞi
þc9 lnðChildren calcium intake=dÞi
þc10 lnðHeightÞit-1 þ u2it ði ¼ 1; 2; .; N; t ¼ 2; 3; 4Þ ð4Þ
Because the observations were available at unequally spaced intervals,
static versions of the models (35) that excluded previous heights were
also estimated for assessing the robustness of the estimated parameters.
The index for childrenÕs total morbidity levels summed up all morbidities
during the 2- to 24-mo period and was included as an explanatory
variable in the models. In addition, data on dietary intakes were compiled at
3 time points, and protein and calcium intakes were averaged to produce
figures for 10–19 y; averaging was helpful in reducing the impact of
within-child variations in intakes (36). In alternative versions of the
model, 2 indicator variables for time periods were included as explan-
atory variables to account for the trends underlying the dietary intakes.
Dynamic and static random-effects models were also estimated for
childrenÕs weights. Last, 4 specifications were estimated for investi-
gating the effects of childrenÕs heights at age 2 y on their completed
heights.
Statistical methods and test statistics. The statistical estimation
Downloaded from https://academic.oup.com/jn/article/146/1/133/4585684 by PYEONGTAEK UNIVERSITY user on 03 October 2020
theory assumed that the number of children (N) was large but the
number of time periods (T) was small. Thus, previous observations on
dependent variables, such as childrenÕs heights and weights, were treated
as correlated with the errors (‘‘endogenous’’) (23). Realizations of certain
time-varying explanatory variables in different survey rounds were
assumed to be uncorrelated with the errors (‘‘exogenous’’). The errors
(u1it) were assumed to be independent across children but correlated over
time with a positive definite variance-covariance matrix. A numerical
optimization routine (E04 JBF) (37) was used in a FORTRAN program
to compute the maximum likelihood estimates. Asymptotic SEs of the
parameters were computed by numerically approximating second
derivatives of the maximized log-likelihood functions. For static models,
stepwise procedures were used to compute the efficient instrumental
variable estimates (35).
Furthermore, certain explanatory variables in the models in Equa-
tions 2 and 4 were likely to be correlated with the errors, especially the
random effects (di). In the model estimated for childrenÕs weights during
ages 2–24 mo, for example, morbidity levels could be correlated with
child-specific random effects. Maximum likelihood methods were used
to consistently estimate the model variables and to test the ‘‘exogeneity’’
null hypothesis for childrenÕs morbidity levels.
Last, in the model explaining childrenÕs completed heights at age
22 y, previous height at age 2 y was likely to be correlated with the error
terms. Specification 1 postulated cross-sectional type regressions that
treated childrenÕs heights at age 2 y as uncorrelated with error terms. In
specification 2, childrenÕs height at age 2 y was treated as correlated with
the error terms in instrumental variable estimation. Specification 3
estimated a dynamic random-effects model by using the data for 2 time
periods and treating height at age 2 y as an exogenous variable. In
specification 4, childrenÕs height at age 2 y was treated as an endogenous
variable in the maximum likelihood estimation.
Results
Table 1 reports sample means and SDs of salient variables used
in the analyses. The sample consisted of a similar proportion of
boys and girls. Mean child birth order was ;3 and number of
persons in the household was 5.6. Maternal highest grade was a
mean of 7.6 and socioeconomic index was 7.25; this index
ranged from 1 to 26. MothersÕ mean height was 150.8 cm and
weight when the child was 1 y was 46.1 kg; the corresponding
mean BMI (in kg/m2) was 20.4. The mean morbidity index for
mothers was 0.58. The mean energy intake was 3144 kcal/d during
the pregnancy and declined to 2333 kcal/d when the child was 1 y.
Mean morbidity indexes 1 and 2, corresponding to the
previous 24 h and 7 d when the children were 1 y, were 0.91 and
1.03, respectively; the respective values at age 2 y were 0.79
and 0.95, which represented a slight decline. The mean number
of child immunizations was 1.38; 63% of the children did
not receive any immunizations. ChildrenÕs heights and weights
showed steady increases. ChildrenÕs energy, protein, calcium,
and b-carotene intakes increased over time. For example, mean
protein and calcium intakes at age 19 y were 70.9 g/d and 520 mg/d,
respectively.
Growth and morbidity patterns of children in Cebu 135
 TABLE 1 Characteristics of children in the Cebu Longitudinal Health and Nutrition Survey from birth to
age 19 y1

Age
Variable Birth 1y 2y 8y 10 y 15 y 19 y

Girls, % 47 — — — — — —
Birth order 3.14 6 2.0 — — — — — —
Number of persons in household 5.57 6 2.8 — — — — — —
Socioeconomic status index 7.25 6 4.0 — — — — — —
MothersÕ — — — — — — —
Highest grade 7.56 6 3.7 — — — — — —

Downloaded from https://academic.oup.com/jn/article/146/1/133/4585684 by PYEONGTAEK UNIVERSITY user on 03 October 2020

Morbidity index — 0.58 6 1.1 0.56 6 1.1 — — — —
Height, cm 150.8 6 5.1 — — — — — —
Weight, kg 50.1 6 6.9 46.1 6 7.3 47.1 6 7.6 52.2 6 9.5 52.7 6 10.0 53.8 6 10.4 55.2 6 10.8
Energy intake, kcal/d 3144 6 1781 2333 6 1521 — — — — —
ChildrenÕs — — — — — — —
Morbidity index 12 — 0.91 6 1.3 0.79 6 1.2 — — — —
Morbidity index 22 — 1.03 6 1.5 0.95 6 1.3 — — — —
Diarrhea index — 0.25 6 1.0 0.18 6 0.8 — — — —
Total number of immunizations3 — — 1.38 6 2.1 — — — —
Height, cm 49.3 6 2.1 70.7 6 3.0 79.1 6 3.7 117.7 6 5.6 133.8 6 7.5 154.0 6 7.8 157.1 6 8.1
Weight, kg 2.99 6 0.5 7.9 6 1.1 9.8 6 1.2 20.6 6 3.1 28.5 6 6.1 44.5 6 7.9 49.8 6 8.7
Energy intake, kcal/d — 434 6 350 727 6 404 — 1201 6 570 1688 6 875 1883 6 1064
Protein intake, g/d — 12.3 6 12.1 21.6 6 15.6 — 38.4 6 19.7 51.6 6 32.2 70.9 6 45.0
Calcium intake, mg/d — 256 6 350 282 6 333 — 322 6 239 364 6 231 520 6 113
b-Carotene intake, μg/d — 132 6 373 350 6 800 — 526 6 906 708 6 1195 758 6 1610
1
Values are means 6 SDs, n = 3080 children.
2
Morbidity index 1 covered the past 24 h, whereas morbidity index 2 covered the past 7 d.
3
Total number of vaccinations received by the age 2 y.

Results for infant lengths and weights at birth. Table 2
presents the results for natural logarithms of lengths and weights
of infants at birth with the use of 2 specifications. The model was
first estimated by using ordinary least squares; ‘‘barangay’’-level
random effects were included in the second specification and the
variables were estimated by maximum likelihood (38, 39). The
results from the 2 models were very close, although a statistical
test indicated that the variance of random effects was statisti-
cally greater than zero. The main findings from these models
were, first, that childrenÕs birth order and number of persons in
the household were associated in a nonlinear fashion with their
lengths and weights at birth. The estimated coefficients showed
that infant lengths and weights significantly (P < 0.05) increased
with birth order, although at a declining rate; lengths and

TABLE 2 Ordinary least-squares and maximum likelihood estimates of models for childrenÕs lengths and weights at birth in Cebu,
Philippines1

Dependent variable
ln (Length), cm ln (Weight), kg
Explanatory variables Ordinary least squares Maximum likelihood2 Ordinary least squares Maximum likelihood2

Constant 3.017 6 0.122 3.028 6 1.171 21.488 6 0.420 21.490 6 0.42

Indicator for girls, 0–1 20.009* 6 0.002 20.009* 6 0.002 20.017* 6 0.005 20.017* 6 0.0004
Birth order 0.005* 0.005* 6 0.001 0.028* 6 0.004 0.028* 6 0.004
(Birth order)2 20.0004* 6 0.0002 20.0004* 6 0.0001 20.002* 6 0.0004 20.002* 6 0.0003
Number of persons in household 20.003* 6 0.001 20.003* 6 0.001 20.008* 6 0.004 20.008* 6 0.004
(Number of persons in household)2 0.0002* 6 0.0001 0.0002* 6 0.0001 0.001* 6 0.0005 0.0005* 6 0.0002
Socioeconomic index 0.001* 6 0.0003 0.001* 6 0.0002 0.003* 6 0.001 0.003* 6 0.001
MothersÕ
Highest grade 0.0005* 6 0.0002 0.0005* 6 0.0003 0.002* 6 0.001 0.001 6 0.001
ln Height, cm 0.130* 6 0.027 0.128* 6 0.027 0.299* 6 0.091 0.301* 6 0.091
ln Weight, kg 0.050* 6 0.007 0.051* 6 0.007 0.252* 6 0.023 0.253* 6 0.023
Energy intake, kcal/d 0.004* 6 0.001 0.004* 6 0.001 0.007 6 0.005 0.005 6 0.005
Chi-square (1) test, multilevel random effects 10.63* 30.07*
R2 0.095* 0.111*
1
Values are regression coefficients 6 SEs, n = 2990 children. *P , 0.05.
2
Multilevel models with random effects for ‘‘barangay.’’

136 Bhargava
weights declined with respect to household size at an increasing
rate. Girls were shorter and lighter than boys at birth.
Second, the coefficients of mothersÕ highest grade and socio-
economic index were positive and significant. This was also
true for mothersÕ heights and weights. The ‘‘elasticity’’ of infant
length with respect to mothersÕ height (percentage change in
infant length resulting from a 1% change in mothersÕ height) was
0.13; the elasticity with respect to mothersÕ weight was 0.05. The
elasticities of infant birth weight with respect to mothersÕ heights
and weights were 0.30 and 0.25, respectively. Third, mothersÕ
energy intakes were significantly associated with infant lengths;
the coefficients were positive but not significant in the model for
the previous time period was also negative and significant.
Although childrenÕs intake of protein expressed as a ratio to
energy intake was not significantly associated, the coefficient of
calcium intake was positive and significantly associated with
childrenÕs heights. The coefficient of the lagged dependent
variable was 0.88, implying large long-run effects of explanatory
variables on childrenÕs heights. For example, whereas the short-
run elasticity of childrenÕs heights with respect to the calcium
to energy intake ratio was 0.001, the long-run elasticity was
;0.01.
The results for childrenÕs weights in Table 3 are presented for
2 specifications—namely, where childrenÕs height was excluded

Downloaded from https://academic.oup.com/jn/article/146/1/133/4585684 by PYEONGTAEK UNIVERSITY user on 03 October 2020

weight.
Results for childrenÕs heights, weights, and morbidity
indexes during ages 2–24 mo. Table 3 presents the results
from dynamic random-effects models for childrenÕs heights and
weights estimated by using data at bimonthly intervals; results
for childrenÕs morbidity indexes are shown in Table 4. In the
model for childrenÕs heights, socioeconomic index and mothersÕ
highest grade were positively and significantly associated,
whereas birth order and number of persons in the household
were negatively associated. MothersÕ BMI and energy intakes
were positively and significantly associated with childrenÕs
heights. By contrast, mothersÕ morbidity index was negatively
associated with childrenÕs heights.
The number of childrenÕs immunizations was positively and
significantly associated with height, although the coefficients
were small. Moreover, coefficients of childrenÕs morbidity
indexes 1 and 2 in the previous time period were negative and
significant and were very close in magnitudes. Thus, the 2
morbidity indexes were combined partly to reduce the number
of variables in the model. The coefficient of the diarrhea index in
from the model and where height was an explanatory variable
(40). In both models, socioeconomic index and mothersÕ highest
grade, BMI, and energy intakes were positively and significantly
associated with weights. By contrast, birth order and number of
persons in the household were negatively associated. Moreover,
childrenÕs number of immunizations and protein to energy
intake ratio were positively and significantly associated with
weights. ChildrenÕs combined morbidity index and the diarrhea
index were estimated with negative coefficients that were
significant. Moreover, there was evidence of correlations be-
tween child-specific random effects (di) and morbidity indexes
and childrenÕs heights. The likelihood ratio statistics for exogeneity
hypotheses in the 2 models for weights were 53.8 and 485.2,
respectively; these statistics were distributed in large samples as
chi-square variables with 33 and 44 df and rejected the null
hypotheses.
The results for childrenÕs morbidity indexes in Table 4 were
similar across the models and underscored the importance of
socioeconomic, environmental, and dietary variables. Girls were
less often sick than boys; socioeconomic index was negatively
and significantly associated with the first 2 morbidity indexes

TABLE 3 Maximum likelihood estimates of dynamic random-effects models for heights and weights of
children in Cebu, Philippines, ages 2–24 mo1

Dependent variable
Explanatory variables ln (Height), cm ln (Weight), kg ln (Weight),2 kg

Constant 0.576 6 0.002 0.527 6 0.022 21.045 6 0.030

Indicator for girls, 0–1 20.002* 6 0.0003 20.009* 6 0.001 20.021* 6 0.001
Socioeconomic index 0.003* 6 0.0004 0.001* 6 0.0001 0.002* 6 0.002
Birth order 20.0004* 6 0.0001 20.001* 6 0.0002 20.001* 6 0.0003
Number of persons in household 20.0002* 6 0.0001 20.001* 6 0.0002 20.001* 6 0.0003
MothersÕ
Highest grade 0.0002* 6 0.0001 0.0004* 6 0.0001 0.001* 6 0.0002
ln BMI 0.006* 6 0.001 0.031* 6 0.003 0.057* 6 0.004
ln Energy intake, kcal/d 0.0005* 6 0.0001 0.002* 6 0.001 0.002* 6 0.001
Morbidity index 20.0002* 6 0.0001 20.0001 6 0.0003 20.0004 6 0.0003
ChildrenÕs
Number of total immunizations 0.0002* 6 0.0001 0.0003* 6 0.0001 0.001* 6 0.0002
Morbidity indexes 1 and 23 20.0003* 6 0.0001 20.003* 6 0.0003 20.004* 6 0.0002
Diarrhea index3 20.0003* 6 0.0001 20.004* 6 0.0003 20.004* 6 0.001
ln protein/energy intake, g/(kcal  d) 0.0004 6 0.0003 0.004* 6 0.001 0.004* 6 0.001
ln calcium/energy intake, mg/(kcal  d) 0.001* 6 0.0001
ln height, cm 0.561* 6 0.004
Lagged dependent variable 0.880* 6 0.0001 0.867* 6 0.003 0.550* 6 0.004
Chi-square test for exogeneity4 (df) 53.8* (33) 485.2* (44)
1
Values are regression coefficients 6 SEs, n = 2076 children repeatedly observed in 11 periods. *P , 0.05.
2
First model for weight did not include childrenÕs height.
3
ChildrenÕs lagged morbidity and nutrient intakes were included in the model for height; current morbidity and intakes were included in the
model for weight.
4
Test if childrenÕs morbidity and diarrhea indexes were correlated with random effects.

Growth and morbidity patterns of children in Cebu 137

TABLE 4 Maximum likelihood estimates of dynamic random-effects models for morbidity indexes of children in Cebu, Philippines,
ages 2–24 mo1

Dependent variable
Explanatory variables ChildrenÕs morbidity index 1 ChildrenÕs morbidity index 2 ChildrenÕs diarrhea index

Constant 2.977 6 0.037 5.516 6 0.040 1.740 6 0.486

Indicator for girls, 0–1 20.068* 6 0.019 20.097* 6 0.019 20.027* 6 0.014
Socioeconomic index 20.008* 6 0.003 0.007* 6 0.002 20.003 6 0.002
Birth order 20.027* 6 0.005 0.004 6 0.005 20.007* 6 0.003
Number of persons in household 0.009* 6 0.004 20.006 6 0.003 0.002 6 0.002
MothersÕ

Downloaded from https://academic.oup.com/jn/article/146/1/133/4585684 by PYEONGTAEK UNIVERSITY user on 03 October 2020

Highest grade 20.014* 6 0.002 20.006* 6 0.003 20.002* 6 0.001
Morbidity index 0.137* 6 0.008 0.062* 6 0.009 0.031* 6 0.008
Household
Piped water, 0–1 20.025 6 0.061 20.037 6 0.069 0.012 6 0.026
Flush toilet, 0–1 20.068 6 0.076 20.039 6 0.087 20.013 6 0.031
General hygiene condition,2 1–4 20.052* 6 0.009 20.049* 6 0.009 0.006 6 0.004
ChildrenÕs
Total number of immunizations 20.015* 6 0.006 0.002 6 0.007 20.002 6 0.002
ln Height, cm 20.388* 6 0.008 20.956* 6 0.008 20.373* 6 0.101
ln b-Carotene/energy intake, μg/(kcal  d) 20.025* 6 0.003 20.009* 6 0.004 20.008* 6 0.003
Lagged dependent variable 0.051* 6 0.008 0.042* 6 0.008 0.582* 6 0.154
Between/within variance ratio 0.086* 6 0.006 0.050* 6 0.005
Chi-square (22) test for exogeneity3 20.9 27.0 42.4*
1
Values are regression coefficients 6 SEs, n = 2065 children repeatedly observed in 11 time periods. *P , 0.05.
2
Higher scores on general hygiene conditions reflect greater cleanliness.
3
Test for whether childrenÕs height and b-carotene to energy intake ratio were correlated with the random effects.

but was not significant in the model for diarrhea index. Whereas
mothersÕ highest grade was significantly negatively associated,
mothersÕ morbidity index was positively associated with all 3
indexes. Moreover, greater hygiene in the house was associated
with lower child morbidity levels. This was also true for number
of immunizations that the children received and for the ratio of
b-carotene to energy intakes.
Finally, the lagged dependent variables were estimated with
small coefficients of 0.05 and 0.04 in the models for morbid-
ity indexes 1 and 2, respectively, and were significant. The co-
efficient of the lagged dependent variable in the model for
diarrhea index was larger (0.58), indicating longer persistence
in the effects of explanatory variables on diarrhea index.
Coefficients of previous heights were negative, indicating that
taller children were sick less often and with lower intensity. The
exogeneity hypothesis for previous heights could not be
rejected in the models for morbidity indexes 1 and 2; the
exogeneity hypothesis was rejected in the model for diarrhea
index.
Results for childrenÕs heights and weights during ages
8–19 y. Table 5 reports the estimates from dynamic and static
random-effects models for childrenÕs heights and weights with
the use of data at 8, 10, 15, and 19 y. First, the coefficient of the
socioeconomic index was significant only in the dynamic model
for childrenÕs heights; this coefficient was positive but was not
significant in the static version of the model. The coefficient of
mothersÕ highest grade was significant in all models except the
dynamic model for weights.
Second, the coefficient of mothersÕ height was estimated to be
large and significant in the models for childrenÕs heights and
weights. Whereas the short-run elasticity of childrenÕs heights
with respect to mothersÕ heights was 0.30 in the dynamic model,
the long-run elasticity was 0.76, which was larger than the
138 Bhargava
corresponding estimate, 0.35, from the static model. In contrast,
the estimated long-run elasticity of childrenÕs weights with
respect to mothersÕ heights was 0.47, which was very close to the
corresponding estimate from the static model.
Third, the number of immunizations received by children
during the ages of 2–24 mo was significantly associated with
height. Moreover, childrenÕs morbidity levels during the ages of
2–24 mo were negatively associated with height and weight in
later years; the coefficients were significant in the static versions
of the models. Last, coefficients of childrenÕs average intakes of
protein and calcium during the ages of 10–19 y were significant
in the model for height; coefficients of protein intakes were
positive and significant in the model for weight. Coefficients of
lagged dependent variables were lower in Table 5 than those in
Table 3 because the models were estimated by using observa-
tions separated by several years.
Results from models for childrenÕs heights at 22 y
explained by current and early variables. Table 6 reports
the results from 4 specifications for childrenÕs completed
heights at 22 y with the use of data on heights at age 2 y. The
results from specification 1 showed that the coefficients of
mothersÕ height and childrenÕs calcium intakes in adoles-
cence were positive and significantly associated with completed
heights at age 22 y. However, the coefficient of calcium intakes
at 2 y was estimated with an unexpected minus sign, which was
significant. The coefficient of height at age 2 y was 0.43, which
was large and significant. In specification 2, estimated by using
instrumental variables, this coefficient decreased to 0.14 and
was no longer significant. The instrumental variables used for
heights at age 2 y were mothersÕ highest grade and weight. A
test for exogeneity (41) showed that there was likely to be a
correlation between errors affecting this model and childrenÕs
heights at age 2 y.
 TABLE 5 Efficient estimates of dynamic and static random-effects models for heights and weights of
children in Cebu, Philippines, by using data at ages 8, 10, 15, and 19 y1

Dependent variable
ln (Height), cm ln (Weight), kg
Explanatory variables Dynamic model2 Static model2 Dynamic model Static model

Constant 1.645 6 0.002 2.904 6 0.147 0.050 6 0.219 20.006 6 0.471

Indicator for girls, 0–1 20.044* 6 0.001 20.025* 6 0.002 20.070* 6 0.004 20.068* 6 0.007
Socioeconomic index 0.0001* 6 0.0001 0.0003 6 0.0003 20.0001 6 0.0004 0.0004 6 0.0009
MothersÕ
Highest grade 0.0003* 6 0.0001 0.001* 6 0.0003 20.0003 6 0.0004 0.002* 6 0.001

Downloaded from https://academic.oup.com/jn/article/146/1/133/4585684 by PYEONGTAEK UNIVERSITY user on 03 October 2020

ln Height, cm 0.298* 6 0.001 0.352* 6 0.030 0.272* 6 0.046 0.474* 6 0.097
ln Weight, kg 20.0003 6 0.002 0.040* 6 0.004 0.055* 6 0.008 0.240* 6 0.015
ChildrenÕs
Total number of immunizations 0.0003* 6 0.0001 0.001* 6 0.0004 0.001* 6 0.0006 0.003* 6 0.001
Total morbidity index 20.0001 6 0.0001 20.0002* 6 0.0001 20.0001 6 0.0002 20.0008* 6 0.0004
ln Protein intake,3 g/d 0.004* 6 0.001 0.009* 6 0.003 0.019* 6 0.004 0.035* 6 0.008
ln Calcium intake,3 mg/d 0.002* 6 0.001 0.005* 6 0.002
Lagged dependent variable 0.391* 6 0.002 0.619* 6 0.008
1
Values are regression coefficients 6 SEs, n = 1700 children repeatedly observed in 4 time periods. *P , 0.05.
2
Dynamic models were estimated by maximum likelihood, whereas static models used a stepwise estimation procedure.
3
Nutrient intakes at ages 10, 15, and 19 y were averaged.

In specification 3, the dynamic model was estimated by using
data at ages 2 and 22 y and by treating height at 2 y as an
exogenous variable. However, coefficients of childrenÕs protein
and calcium intakes were not statistically different from zero. In
specification 4, which treated previous height as an endogenous
variable, the estimated coefficients of protein and calcium
intakes were 0.006 and 0.003, respectively, and were significant.
The likelihood ratio test statistic for the exogeneity hypothesis
for heights at 2 y was 13.6 and indicated that child-specific
random effects were correlated with previous heights. Moreover,
the coefficient of childrenÕs heights at 2 y decreased to 0.025
when height was treated as an endogenous variable and remained
significant.
Discussion
This article presented 4 sets of comprehensive analyses for
growth and morbidity patterns of children in Cebu, Philippines,
and tackled the conceptual and methodologic aspects. First, the
results for infant lengths and weights at birth underscored the
importance of mothersÕ nutritional status reflected in height,
weight, and energy intakes. Moreover, in alternative models for

TABLE 6 Efficient estimates of models for completed heights of children in Cebu, Philippines, at age 22 y
explained by variables including heights at age 2 y1

Dependent variable (ln height at age 22 y): estimation method, cm

Explanatory variables Specification 12 Specification 23 Specification 34 Specification 45

Constant 1.609 6 0.107 2.341 6 0.250 1.704 6 0.024 2.671 6 0.001

Indicator for girls, 0–1 20.067* 6 0.001 20.071* 6 0.002 20.067* 6 0.024 20.073* 6 0.001
Socioeconomic index 0.0001 6 0.001 0.0002 6 0.0002 20.0001 6 0.024 0.0004* 6 0.0001
MothersÕ ln height, cm 0.331* 6 0.021 0.439* 6 0.041 0.320* 6 0.024 0.471* 6 0.001
ChildrenÕs ln
Protein intakes at age 2 y, g/d 0.001 6 0.002 0.001 6 0.001
Calcium intakes at age 2 y, mg/d 20.003* 6 0.001 20.0016 6 0.0014
Protein intake in adolescence,6 g/d 0.0004 6 0.002 0.0041 6 0.003
Calcium intake in adolescence, mg/d 0.002* 6 0.002 0.0024 6 0.002
Time-varying protein intake, g/d 20.002 6 0.024 0.006* 6 0.0002
Time-varying calcium intake, mg/d 0.002 6 0.023 0.003* 6 0.001
Height at age 2 y, cm 0.431* 6 0.016 0.137 6 0.092 0.424* 6 0.024 0.025* 6 0.0002
Chi-square (1) test for overidentification 0.08
Chi-square (1) test exogeneity height at age 2 y 13.1* 13.6*
1
Values are regression coefficients 6 SEs, n = 1692 children repeatedly observed in 2 time periods. *P , 0.05.
2
ChildrenÕs height at age 2 y was treated as an exogenous variable and the model was estimated by ordinary least squares.
3
ChildrenÕs height at age 2 y was treated as an endogenous variable with mothersÕ highest grade and weight used as instrumental
variables.
4
ChildrenÕs height at age 2 y was treated as an exogenous variable and the model was estimated by maximum likelihood.
5
ChildrenÕs height at age 2 y treated as an endogenous variable and the model was estimated by maximum likelihood.
6
Nutrient intakes at ages 10, 15, and 19 y were averaged.

Growth and morbidity patterns of children in Cebu 139

infant birth lengths and weights, mothersÕ energy intakes from
fat were estimated with coefficients 0.004 and 0.007, respec-
tively; the coefficient was significant in the model for length. It is
often difficult to assess the effects of mothersÕ diet quality on
birth outcomes, partly because limited information is avail-
able on dietary intakes during pregnancy trimesters (31). Future
research compiling data on mothersÕ food intakes during
pregnancy at regular intervals could shed light on these issues.
Second, the analyses of childrenÕs growth and morbidity
patterns during the ages of 2–24 mo provided several insights.
Although it is recognized that catch-up growth in developing
countries critically depends on morbidity levels (42), children
are often sick with multiple illnesses, which exacerbate nutrient
loss and are difficult to account for in empirical models. It was
therefore important to construct 3 indexes of child morbidity
covering the duration and intensity. The results showed that the
morbidity indexes were negatively and significantly associated
with childrenÕs heights and weights. Thus, greater resources are
necessary for improving hygiene in the households, which was
associated with lower child morbidity levels; this will, in turn,
enhance physical growth. Furthermore, childrenÕs intakes of
calcium, protein, and b-carotene were significantly associated
with height, weight, and morbidity indexes, respectively. Thus,
nutritional interventions to increase the intakes of animal
products and fruit and vegetables are likely to synergistically
enhance child growth, especially when the factors underlying
sicknesses are simultaneously tackled (1).
Third, the data on childrenÕs dietary intakes and anthropo-
metric measures at ages 8, 10, 15, and 19 y were useful for
analyzing the dynamics of growth in adolescence. ChildrenÕs
protein and calcium intakes were significantly associated with
height; protein intakes were significantly associated with weight.
This evidence is helpful for reconciling contrasting views on the
extent of catch-up growth in developing countries. For example,
using logistic models but not controlling for between-children
differences, previous researchers (43) reported catch-up growth
among Cebu children ages 2–9 y. More recently, these researchers
(44) questioned the efficacy of interventions for children ‘‘after
1000 days.’’ The present analyses of Cebu data spanning several
years underscored the importance of nutritional intakes during
early childhood as well as in adolescence. Longitudinal studies
covering several years can provide insights that are as informa-
tive as the evidence from randomized controlled trials, which
typically cover shorter periods and may revisit children after
long intervals (45). This is partly because, in the broader
anthropologic context, nutritional interventions can improve
childrenÕs ‘‘mixed’’ diets and nutrient absorption rates to a limited
extent.
Finally, for comparisons with the previous literature, models
were estimated for childrenÕs completed heights at age 22 y by
using explanatory variables from when the children were 2 y old.
Previous estimation methods were inadequate for tackling the
dynamics of child growth and the correlations between previous
heights and the error terms affecting current heights. The
seemingly large magnitudes of the estimated coefficient of height
at 2 y should not be interpreted as reflecting inflexibility in
childrenÕs growth patterns. In fact, treating the heights at 2 y
as correlated with the error terms led to smaller estimated co-
efficients and significant coefficients of protein and calcium
intakes. Thus, depending on the prevalence of stunting, mor-
bidity environments, and policy goals, nutritional interventions
can supply the deficient nutrients over a longer time frame than
just in the first 2 y. Food policy interventions typically improve
the quality of mixed diets of older children, thereby facilitating
140 Bhargava
linear growth, which, in turn, determines their adult physical
work capacity and labor productivity.
Acknowledgments
While retaining responsibility for the views, I thank L Adair,
A Bas, E Birol, H Bouis, L Howard, I Larenas, N Lee, and
M Zeller for their valuable help. This article is dedicated to the
memory of Nevin S Scrimshaw, who selflessly spent his valuable
time educating AB regarding the nutrition-infection interac-
tions. AB had responsibility for all parts of the manuscript and
read and approved the final manuscript.
Downloaded from https://academic.oup.com/jn/article/146/1/133/4585684 by PYEONGTAEK UNIVERSITY user on 03 October 2020
References
1. Scrimshaw NS, Taylor CE, Gordon JE. Interactions of nutrition and
infection. Am J Med Sci 1959;237:367–403.
2. Scrimshaw NS, Behar M. Protein malnutrition in young children.
Science 1961;133:2039–47.
3. Scrimshaw NS. Nutrition and health from womb to tomb. Nutr Today
1996;31:55–67.
4. Scrimshaw NS, SanGiovanni J. Synergism of nutrition, infection, and
immunity: an overview. Am J Clin Nutr 1997;66:464S–77S.
5. Keusch G. Nutrition and infection. In: Shils M, Olson J, Shike M,
editors. Modern nutrition in health and disease. Philadelphia: Lea and
Febiger; 1994. p. 1241–58.
6. Calder PC. Feeding the immune system. Proc Nutr Soc 2013;72:299–309.
7. Gibson RS, Bailey K, Gibbs M, Ferguson E. A review of phytate, zinc,
and calcium concentrations in plant-based complementary foods used in
low-income countries and implications for bioavailability. Food Nutr
Bull 2010;31:S134–6.
8. Bhargava A, Bouis HE, Scrimshaw NS. Dietary intakes and socioeco-
nomic factors are associated with the hemoglobin concentration of
Bangladeshi women. J Nutr 2001;131:758–64.
9. Mata L. The children of Santa Maria Cauque. Cambridge (MA): MIT
Press; 1978.
10. Spurr GB. Nutritional status and physical work capacity. Am J Phys
Anthropol 1983;26:1–35.
11. Bhargava A. Nutritional status and the allocation of time in Rwandese
households. J Econom 1997;77:277–95.
12. Bhandari N, Bahl R, Nayyar B, Khokhar P, Rohde J, Bhan M. Food
supplementation with encouragement to feed it to infants from 4 to 12
months of age has a small impact on weight gain. J Nutr
2001;131:1946–51.
13. Tanner JM, Whitehouse R, Takaishi M. Standards from birth to
maturity for height, weight, height velocity, and weight velocity. Arch
Dis Child 1966;41:454–71.
14. Tanner J. Foetus into man. Cambridge (MA): Harvard University Press;
1990.
15. Martorell R, Khan L, Schroeder D. Reversibility of stunting: epidemi-
ological findings in children from developing countries. Eur J Clin Nutr
1994;48:S45–57.
16. Prentice AM, Ward K, Goldberg G, Jarjou L, Moore S, Fulford A,
Prentice A. Critical windows for nutritional interventions against
stunting. Am J Clin Nutr 2013;97:911–8.
17. Bhargava A. Diet quality, child health and food policies in developing
countries. World Bank Res Obs 2015;30:247–76.
18. Schönbeck Y, Talma H, van Dommelen P, Bakker B, Buitendijk S,
HiraSing R, van Buuren S. The worldÕs tallest nation has stopped
growing taller: the height of Dutch children from 1955–2009. Pediatr
Res 2013;73:371–7.
19. Günther AL, Karaolis-Danckert N, Remer T, Buyken A. Dietary protein
intake throughout childhood is associated with the timing of puberty.
J Nutr 2010;140:565–71.
20. Van Vught AJ, Dagnelie P, Arts I, Froberg K, Andersen L, El-Naaman B,
Bugge A, Nielsen B, Heitman B. Dietary arginine and linear growth:
the Copenhagen School Child Intervention Study. Br J Nutr 2013;109:
1031–9.
21. Sørensen HT, Sabroe S, Rothman K, Gillamn M, Steffensen F, Fischer P,
Sorensen T. Birth weight and length as predictors of adult heights. Am J
Epidemiol 1999;149:726–9.
22. Adair LS, Popkin B, Akin J, Guilkey D, Gultiano S, Borja J, Perez L,
Kuzawa C, McDade T, Hindin M. Cohort profile: the Cebu Longitu-
dinal Health and Nutrition Survey. Int J Epidemiol 2011;40:619–25.
23. Bhargava A, Sargan JD. Estimating dynamic random effects models
from panel data covering short time periods. Econometrica 1983;51:
1635–59.
24. Allen LH, Wood RJ. Calcium and phosphorus. In: Shils ME, Olson JA,
Shike M, editors. Modern nutrition in health and disease. 8th ed.
Philadelphia: Lea and Febiger; 1994.
25. Rand Corporation. Measurement of physiologic health for children.
Vol. 1–5. Santa Monica (CA): Rand Corporation; 1983.
26. Bhargava A. Modelling the health of Filipino children. J R Stat Soc Ser
A Stat Soc 1994;157:417–32.
27. Sommer A, West K. Vitamin A deficiency. New York: Oxford University
Press; 1996.
28. Food and Nutrition Research Institute. Food composition tables
recommended for use in the Philippines. Handbook 1, 6th revision.
Manila (Philippines): Department of Science and Technology; 1990.
29. Kronmal R. Spurious correlation and the fallacy of the ratio standard
revisited. J R Stat Soc Ser A Stat Soc 1993;156:379–92.
30. Cole T. Weight-stature indices to measure under-weight, over-weight
and obesity. In: Himes J, editor. Anthropometric assessment of nutritional
status. New York: Wiley-Liss; 1991. p. 83–111.
31. Falkner F, Holzgreve W, Schloo R. Prenatal influences on postnatal
growth: overview and pointers for research. Eur J Clin Nutr 1994;48:
S15–22.
32. Laird NM, Ware JH. Random effects models for longitudinal data.
Biometrics 1982;38:963–74.
33. FAO/UNU/WHO. Energy and protein requirements. Geneva (Switzer-
land);WHO; 1985. (Technical Report 724.)
34. Bhargava A, Reeds PJ. Requirements for what? Is the measurement of
energy expenditure a sufficient estimate of energy needs? J Nutr
1995;125:1358–62.
35. Bhargava A. Identification and panel data models with endogenous
regressors. Rev Econ Stud 1991;58:129–40.
36. Nelson M, Black AE, Morris JA, Cole TJ. Between-and-within subject
variation in nutrient intake from infancy to old age: estimating the
number of days to rank dietary intakes with desired precision. Am J Clin
Nutr 1989;50:155–67.
37. Numerical Algorithm Group. Numerical Algorithm Group mark 13.
Oxford (United Kingdom): Oxford University; 1991.
38. Goldstein H. Multilevel statistical models. New York: Oxford Univer-
sity Press; 1995.
Downloaded from https://academic.oup.com/jn/article/146/1/133/4585684 by PYEONGTAEK UNIVERSITY user on 03 October 2020
39. StataCorp. Stata version 13. College Station (TX): StataCorp; 2013.
40. Ehrenberg A. The elements of lawlike relationship. J R Stat Soc Ser A
Stat Soc 1968;131:280–302.
41. Rao CR. Large sample tests of statistical hypotheses concerning several
parameters with applications to problems of estimation. Proc Camb
Philos Soc 1948;44:50–7.
42. Golden MHN. Is complete catch-up growth possible for stunted
malnourished children? Eur J Clin Nutr 1994;48:S58–70.
43. Adair LS. Filipino children exhibit catch-up growth from age 2–12
years. J Nutr 1999;129:1140–8.
44. Adair LS, Fall C, Osmond C, Stein A, Martorell R, Ramirez-Zea M,
Sachdev H, Dahley D, Bas I, Norris S, et al; COHORTS Group.
Associations of linear growth and relative weight gain during early life
with adult health and human capital in countries of low and middle
incomes: findings from five birth cohort studies. Lancet 2013;382:525–34.
45. Bhargava A. Food, economics, and health. Oxford (United Kingdom):
Oxford University Press; 2008.
Growth and morbidity patterns of children in Cebu 141