Biology Reference
In-Depth Information
spread of smallpox (Blower and Bernoulli 2006). Since then, a broad spectrum
of modeling frameworks has been developed, with statistical to mechanistic
approaches and aims and different purposes for each method. A mathemat-
ical model is most commonly used for predicting disease risk (incidence)
given previous data on the disease cases. As we have previously discussed
in this chapter, infectious disease transmissions are influenced by environ-
mental and ecological factors, which can be characterized by remote sensing
data. One can use remote sensing data in a mathematical model as the predic-
tors for disease risk. For indirectly transmitted diseases such as vectorborne
ones, remote sensing data can be used in a model to predict the geographic
distribution and abundance of the vector and disease prevalence.
Often, direct relationships between disease transmission and environ-
mental and ecological factors are not clearly understood. Typically, this
calls for a statistical modeling to predict disease risk. The widely used
statistical model is generally a correlation-based approach that does not
require extensive knowledge on the biology of the disease. On the other
hand, when sufficient detail on the biology of the disease is known, one
may use biological-based modeling, such as the compartmental Susceptible
Exposed Infected Recovered (SEIR) Model. Such models as the SEIR model
often not only offer information for risk prediction but also provide an
understanding of the quantitative nature of disease transmission. We will
discuss both statistical and biological models that are commonly used in
the following sections.
3.5.1 Statistical Modeling
Statistical-based modeling generally uses a correlative approach to infer the
empirical relationship between the independent and the dependent variables.
Since we are dealing here with modeling techniques using remote sensing
data, the independent variables, or predictors, in this chapter typically refer
to environmental factors. The dependent variables, or outcome, can be the
disease incidence or the presence (or absence) of disease.
Again, statistically based models are popular due to their ease of use as
they do not require detailed knowledge about disease transmission mecha-
nisms. Although models of this type have good predictive power, it should
be noted that statistical models are based on correlation from past data. This
means that the model assumes no evolution in the transmission dynamics.
Hence, any nonlinearity in the transmission process, which may be embed-
ded in the data, cannot be captured in the model.
3.5.1.1 Logistic Regression
Logistic regression falls under the category of generalized linear models,
and it is the most common statistical technique for predicting the probability
of disease occurrence using environmental variables. The method does not
Search WWH ::




Custom Search