Agriculture Reference
In-Depth Information
finally, showed the use of model parameter estimates in comparing treatment
effects on disease development.
This chapter aims to provide brief introductions to the various quantitative
tools used to describe, analyse and compare temporal dynamics of epidemics.
First, main methods for summarising or modelling each individual observed
epidemic are described. Then, several methods aiming to describe observed
temporal epidemic patterns in fewer dimensions (often 2-3) are briefly introduced;
these methods are based on similarity or dissimilarity between observed/derived
data describing temporal epidemic development over many epidemics or
treatments. Finally, a brief introduction is given to several methods for comparing
epidemic development between treatments using either original data and/or those
derived from the original data, such as estimates of model parameters and newly
transformed data in new dimensions. In this chapter, treatments are defined in a
very broad sense. This may include epidemics at different sites, or on different
cultivars, or subjected to different management programmes; it may also mean
different pathosystems altogether.
8.2 GENERAL CONSIDERATIONS
The degree of precision and complexity of the required modelling and analysis is
determined by the question(s) researchers are trying to answer. Each analytical
method is best suited to answer specific questions. Even for the same data set,
different methods may be needed to answer different questions. For example, if
we have collated disease development on several occasions for one cultivar that
received different fungicide spray regimes, we may simply apply analysis of
variance to the final disease incidence/severity or to the areas under disease
progress curves to determine whether overall disease development is significantly
affected by the spray regimes. To answer the question on how temporal dynamics
are affected by the fungicide spray programme, we may need to examine and
summarise the pattern of disease progress over time for each individual spray
programme and compare the patterns. We could fit growth curve models to each
epidemic data set and then use analysis of variance to determine which and how
each growth curve model parameter is affected by the spray treatment.
Alternatively, we could apply canonical variate analysis to the original or derived
(such as growth model parameters) data from which we can interpret the
differences in observed temporal patterns between treatment groups in relation to
each individual disease assessment, provided there are several replicate epidemic
data sets for each treatment. There are several other methods that could be used to
answer the same question. Precisely which one to use depends on many factors,
including the temporal pattern of the observed epidemics, the nature of data and
their collation, and availability of statistical software packages. In essence, there
are no fixed prescriptions as to which method is the best for every possible
scenario. Researchers need to consider many factors before applying one or more
particular methods to the observed data in order to answer the predefined
questions.
