Analyses
We analyzed the data both descriptively and using multivariate
analyses. For descriptive analyses, we used both the
chi-square test and the student’s t-test. We also used Fisher’s
exact test to describe years of nutrition policies, using
10 years as a cut-off, in LIC and MICs. The 10 years was
arbitrary chosen based on the average time taken to update
a policy as observed in GINA. To examine the presence
of nutrition policies in LICs and MICs, we used the
chi-square test. We compared presence of a policy that
address undernutrition and overweight separately, stratified
by the country’s income level. We used the student’s
t-test to compare mean rates of stunting, wasting, underweight,
and overweight by country’s level of income.
We conducted regression analyses using mixed effects
models with random-intercept by country, to take into account
differences each country has. We built four models
using each type of nutrition status as an outcome variable.
These were stunting, wasting, underweight, and overweight.
We had two main independent variables; nutrition
policy and nutrition governance. For nutrition policy, we
coded as 1 if a country had nutrition policy to address undernutrition
or overweight that was in effect at each anthropometric
measurement year. We used undernutrition
policy for models that had undernutrition as an outcome
variable and overweight policy for model with overweight
status as the outcome variable. A similar method was used
to generate nutrition governance variable that was in effect
at time of anthropometric data collection.
The main confounding variables included GDP per
capita, net percentage of primary school enrolment, and
country governance. We matched years of data of GDP
per capita and net school enrolment with years of anthropometric
measurements to control for the potential
effect each had on the outcome variables. We also controlled
for the year of anthropometric measurement to
observe changes of the outcome variables with time.
We set the statistical significance at p-value < 0.05 and
conducted all analyses using STATA version 12.
Analyses
We analyzed the data both descriptively and using multivariate
analyses. For descriptive analyses, we used both the
chi-square test and the student’s t-test. We also used Fisher’s
exact test to describe years of nutrition policies, using
10 years as a cut-off, in LIC and MICs. The 10 years was
arbitrary chosen based on the average time taken to update
a policy as observed in GINA. To examine the presence
of nutrition policies in LICs and MICs, we used the
chi-square test. We compared presence of a policy that
address undernutrition and overweight separately, stratified
by the country’s income level. We used the student’s
t-test to compare mean rates of stunting, wasting, underweight,
and overweight by country’s level of income.
We conducted regression analyses using mixed effects
models with random-intercept by country, to take into account
differences each country has. We built four models
using each type of nutrition status as an outcome variable.
These were stunting, wasting, underweight, and overweight.
We had two main independent variables; nutrition
policy and nutrition governance. For nutrition policy, we
coded as 1 if a country had nutrition policy to address undernutrition
or overweight that was in effect at each anthropometric
measurement year. We used undernutrition
policy for models that had undernutrition as an outcome
variable and overweight policy for model with overweight
status as the outcome variable. A similar method was used
to generate nutrition governance variable that was in effect
at time of anthropometric data collection.
The main confounding variables included GDP per
capita, net percentage of primary school enrolment, and
country governance. We matched years of data of GDP
per capita and net school enrolment with years of anthropometric
measurements to control for the potential
effect each had on the outcome variables. We also controlled
for the year of anthropometric measurement to
observe changes of the outcome variables with time.
We set the statistical significance at p-value < 0.05 and
conducted all analyses using STATA version 12.
การแปล กรุณารอสักครู่..
