Adult obesity is an important public health issue in the United
States and varies considerably across geographic areas. While obesity is
frequently examined at the national or state level, substantial
differences can also occur among individual counties. Counties differ in
their socioeconomic conditions, physical activity patterns, food
security, behavioral characteristics, and environmental conditions.
Many of the relationships between adult obesity and factors such as
physical inactivity, socioeconomic conditions, and food-related
conditions are already well established. Therefore, the primary purpose
of this project is not simply to determine whether these known factors
are associated with obesity. Instead, this study will examine whether
geographic differences among U.S. counties remain after accounting for
these known county-level characteristics. The project will investigate
the extent to which county-level geographic variation in adult obesity
prevalence can be explained by behavioral, socioeconomic, food-related,
and environmental characteristics. The analysis will first estimate the
variation in obesity prevalence across counties and then examine whether
substantial geographic differences remain after the known factors are
included in a multivariable model.
Because the study uses
observational county-level data, the purpose is to identify statistical
associations rather than causal relationships. The analysis will not
determine whether living in a particular county causes obesity. Instead,
it will examine whether differences between counties remain
statistically meaningful after controlling for measured county-level
characteristics.
Understanding geographic differences in obesity is important because
health outcomes can vary substantially across communities. Two counties
may have similar levels of physical activity, socioeconomic resources,
food insecurity, and environmental characteristics but still have
different adult obesity prevalence. Examining this remaining geographic
variation may provide additional information about health inequality at
the county level.
The study may also demonstrate how quantitative
regression methods can be used to investigate geographic variation in a
public-health outcome using data from multiple sources. Rather than
focusing only on whether individual predictors are statistically
significant, the analysis will compare models, evaluate explained
variation, examine effect sizes and uncertainty, and assess regression
assumptions.
The findings will be interpreted at the county level
and will not be used to make individual-level conclusions. Geographic
differences that remain after adjustment may indicate that other
county-level characteristics, spatial patterns, or unmeasured factors
warrant further investigation.
Obesity and Geographic Differences
Previous public-health research has established adult obesity as an
important population-health issue and has examined relationships between
obesity and behavioral, socioeconomic, and environmental
characteristics. Much of the available discussion focuses on national or
state-level patterns. County-level analysis provides a more detailed
geographic perspective because communities within the same state can
have substantially different health and socioeconomic characteristics.
The county-level perspective is particularly useful for this
project because the research question is concerned with whether
differences between geographic areas remain after accounting for
measured risk factors. Rather than treating county location only as a
source of descriptive variation, the analysis will investigate whether
geographic differences persist in adjusted models.
Known Factors
Associated with Adult Obesity
Physical activity is an important
behavioral characteristic when examining obesity, and differences in
physical activity across communities may be associated with differences
in obesity prevalence. The Institute for Health Metrics and Evaluation
(IHME) provides the United States Physical Activity and Obesity
Prevalence, County dataset, which contains county-level measures of
physical activity and obesity prevalence.
Socioeconomic
characteristics such as education, employment, and economic resources
are also relevant county-level characteristics. Counties with different
socioeconomic conditions may have different access to health resources,
food, and opportunities for physical activity. Food insecurity may also
be related to socioeconomic conditions and therefore needs to be
considered together with other socioeconomic variables.
These
factors will be treated primarily as known explanatory or control
variables in the proposed analysis. The purpose is not to establish that
these relationships exist, but to determine whether county-level
geographic differences in obesity remain after accounting for them.
Geographic and Spatial Variation
The central research issue
is whether the location of a county is associated with differences in
adult obesity prevalence beyond the measured characteristics of that
county. If geographic differences remain after adjustment, the remaining
variation could reflect other county-level conditions, spatially
patterned factors, or variables that are not included in the available
datasets. The project will therefore compare an adjusted model
containing the known risk factors with models that incorporate
geographic or spatial information, when the available data and project
scope support such an analysis. This approach shifts the emphasis from
simply identifying known obesity correlates to measuring the amount of
geographic variation that remains after adjustment.
Originality
of the Research
The originality of this project does not depend
on claiming that the relationship between obesity and physical activity,
socioeconomic conditions, food insecurity, or environmental
characteristics has never been studied. Instead, the project will focus
on a more specific question: whether meaningful county-level geographic
differences in adult obesity prevalence remain after these known factors
are controlled for. The analysis will combine county-level measures of
adult obesity, physical activity, education, unemployment, wealth, food
insecurity, and other available characteristics within a common
regression framework. The project will compare models and evaluate
whether geographic or spatial information provides additional
explanatory value after adjustment for the known factors. This approach
provides a quantitative way to examine geographic health inequality
while recognizing that the study is observational and that unmeasured
factors may contribute to the remaining county-level variation.
To what extent do county-level geographic differences in adult obesity prevalence remain after controlling for known behavioral, socioeconomic, food-related, and environmental factors?
Supporting Research Questions
1. How much variation in adult
obesity prevalence exists across U.S. counties before accounting for
other county-level characteristics?
2. To what extent do known
behavioral, socioeconomic, food-related, and environmental
characteristics explain differences in adult obesity prevalence across
counties?
3. After controlling for known county-level factors, do
substantial differences in adult obesity prevalence remain across U.S.
counties?
4. Does geographic or spatial information provide
additional explanatory information about county-level adult obesity
prevalence beyond the known risk factors?
5. How does model
performance change when geographic or spatial information is considered
in addition to the known county-level risk factors?
H1: Adult obesity prevalence varies significantly across U.S.
counties.
H2: County-level behavioral, socioeconomic, food-related,
and environmental characteristics explain a significant portion of the
variation in adult obesity prevalence across counties.
H3: After
controlling for known behavioral, socioeconomic, food-related, and
environmental characteristics, significant geographic differences in
adult obesity prevalence remain across U.S. counties.
H4: A model
that incorporates geographic or spatial information provides additional
explanatory information about county-level adult obesity prevalence
beyond a model containing only the known county-level risk factors.
Research Methods, Models, Variables, and Data Sets
Research Design The study will use a quantitative, observational research design. The primary unit of analysis will be the U.S. county. Data from multiple publicly available datasets will be combined to create a county-level analytical dataset. County identifiers, including Federal Information Processing Standards (FIPS) codes where available, will be used to merge the datasets. The analysis will use a sequential modeling strategy. First, the amount of geographic variation in adult obesity prevalence will be described. Second, a multivariable model will be estimated using known behavioral, socioeconomic, food-related, and environmental characteristics. Third, geographic or spatial information will be considered to determine whether meaningful geographic variation remains after adjustment. Because the data are observational and aggregated at the county level, the results will not be interpreted as individual-level causal relationships. The analysis will describe county-level associations and geographic variation. Modeling Approach The initial analysis will describe the distribution of adult obesity prevalence across counties and measure the amount of between-county variation. The first regression model will include the known county-level factors. This model will provide a baseline for determining how much of the observed variation in adult obesity prevalence is associated with measured behavioral, socioeconomic, food-related, and environmental characteristics. A second model will incorporate geographic or spatial information, if the final data structure and project scope support an appropriate spatial specification. The two models will be compared using measures such as explained variation, model fit, prediction error, and statistical uncertainty. The main analytical question will be whether meaningful geographic variation remains after the known risk factors are controlled for. Multicollinearity, residual behavior, influential observations, and other relevant regression assumptions will be assessed before interpreting the final model. Dependent Variable The primary dependent variable will be: Adult obesity prevalence The primary obesity measure will come from the IHME United States Physical Activity and Obesity Prevalence, County dataset. The final study year will be selected after examining the years available in the obesity dataset and the other predictor datasets. Preference will be given to a year or period for which the major variables are sufficiently comparable. Independent and Control Variables The potential explanatory and control variables will include the following categories: Behavioral factors • Physical activity or physical inactivity • Alcohol consumption Socioeconomic factors • Education • Unemployment • Wealth Food-related factors • Food insecurity Environmental factors • Sunlight or other relevant environmental measures available in the data Population/contextual information • County population Population will primarily be used as contextual information and will also be considered when evaluating whether weighted regression is appropriate. Population will not automatically be used as a regression weight simply because counties differ in size. Data Sets IHME United States Physical Activity and Obesity Prevalence, County This dataset will provide the primary adult obesity prevalence measure and physical activity measure. Education.csv This dataset will provide a county-level education measure. Unemployment.csv This dataset will provide a county-level unemployment measure. Wealth.csv This dataset will provide a county-level measure of wealth or economic resources. FoodInsecurity.csv This dataset will provide a county-level measure of food insecurity. AlcoholConsumption.csv This dataset will provide a county-level behavioral measure related to alcohol consumption. Sunlight.csv This dataset will provide a county-level environmental measure related to sunlight exposure. population.csv This dataset will provide county population information. Additional datasets, including IHME life expectancy and cardiovascular disease measures, may be examined for exploratory analysis. These variables will not automatically be included in the primary model because their relevance and time periods need to be evaluated. Data Preparation The datasets will first be examined individually before they are merged. Variable definitions, units, years, missing values, and county identifiers will be reviewed for comparability. The datasets will then be merged using county identifiers, preferably FIPS codes. Counties with missing values in variables required for the primary analysis will be evaluated and handled using an approach appropriate for the amount and pattern of missingness. Descriptive statistics and visualizations will be used to examine the distribution of obesity prevalence and the potential explanatory variables. Correlations and multicollinearity diagnostics will be evaluated before fitting the final multivariable models. The final analytical dataset and model specification will depend on the availability and comparability of the source datasets. Any changes to the planned variables will be documented in the final report.
Measurement and Evaluation The primary outcome will be county-level adult obesity prevalence. The primary explanatory framework will distinguish between known county-level risk factors and remaining geographic variation. The analysis will evaluate descriptive differences in obesity prevalence across counties, coefficient estimates and uncertainty for the known factors, overall model fit, explained variation, and prediction error. Residual diagnostics and other regression assumptions will also be assessed. The key comparison will be between a model containing known behavioral, socioeconomic, food-related, and environmental characteristics and a model that additionally accounts for geographic or spatial information when appropriate. Evidence that geographic variation remains after adjustment will be evaluated from the statistical significance and magnitude of the geographic component, changes in model fit, and the amount of variation explained. Because this is an observational county-level study, results will be described as associations. The analysis will not claim that geographic location itself causes adult obesity, and county-level findings will not be interpreted as individual-level effects.
Validation of the Tidy Dataset
Vaccination Coverage Countries
Vaccination Progress Over Time
Dataset 2 — Renewable Energy Capacity Time Series
Data Loading and Initial Inspection
Total Installed
Renewable Energy Capacity
Total Renewable Energy
Capacity
Interpretation and Key
Findings
Dataset 3 — GDP by Country (1960–2022)
Data Loading and Initial Inspection
Validation of the Tidy Dataset
GDP Trends for Selected Countries
Average GDP for Selected Countries