Working with Models

Source: vignettes/Working.Rmd 
#devtools::load_all()

ramp.xds was designed to reduce the costs of model building and computation for malaria and other mosquito-transmitted pathogens. One of the core challenges for the software was to provide users with a set of utilities that would make it easy to do work with models after they’ve been built.

Set Up

The function xds_setup was discussed in Getting Started.

mod <- xds_setup(Xname = "SIS", MYZname = "SI", Lname = "basicL")
mod$MYZname
## [1] "SI"

Solving

The functions xds_solve can be called to solve dynamical systems. In these systems, time is the independent variable. Given a set of initial values, xds_solve or dts_solve compute the values of the dependent variables at those time points. Other functions return stable orbits for models with a canonical seasonal signal.

mod <- xds_solve(mod)

Outputs

The variables and terms are returned as raw and parsed outputs:

  • raw outputs are returned in tabular form

  • outputs are parsed into named lists; the values of the dependent variables can be retrieved by name.

  • dynamical terms are computed and returned.

names(mod$outputs)
## [1] "time"   "last_y" "orbits" "terms"  "deout"

Visualize

a family of functions makes it easy to plot basic outputs of the models.

Initial Values

A set of utilities was developed to work with initial values, making it easy to change the intitial values through a function call. In solving, xds_solve calls get_inits with a flatten=TRUE

get_inits(mod, flatten=TRUE)
## [1]   1   5   1 999   1

With the default, flatten=FALSE, get_inits returns a named list that is easy to read:

## $L
## $L$L
## [1] 1
## 
## 
## $MYZ
## $MYZ$M
## [1] 5
## 
## $MYZ$Y
## [1] 1
## 
## 
## $X
## $X$S
## [1] 999
## 
## $X$I
## [1] 1

Since it is a named list, it’s easy to examine a specific element in that list:

get_inits(mod)$MYZ
## $M
## [1] 5
## 
## $Y
## [1] 1

Parameters

Another set of utilities was developed to change parameter values through a function call.

Analyze

The software includes functions that perform various analytical tasks:

  • Compute the steady states of model components using a method that does not involve simulation. These functions also provide a way of independently verifying the code.

  • Compute stable orbits.

  • Compute vectorial capacity and its spatial-temporal dispersion, the human transmitting capacity and its spatial-temporal dispersion, and malaria reproductive numbers as pseudo-threshold conditions.

Modifying Models

ramp.xds has built-in utilities to make it easy to work with the models, including functions to visualize the outputs, to examine and change initial values and parameters, and to analyze the models.

The xds model object – any object that can be solved by xds_solve or dts_solve – is a compound list in R. If you’re already familiar with the format of the xds model object, it’s possible examine any part of it, or change any of the parameter values or initial conditions. While that might be the way to go in the long run, the xds model object includes a set of utilities to retrieve initial values, parameters, and some other values of interests.

The code and documentation for these utilities is designed to be instructive, and a goal for ramp.xds has been to provide detailed documentation, include references to peer reviewed journal articles.

Some of the values affect the rigid interfaces for egg laying and emergence and for blood feeding and transmission, so changing the values without updating the interface would create a model that was internally inconsistent. To handle this, ramp.xds has introduced functions to make it easy to change values after setup.

From the user perspective, using these change_ functions will make it easy to avoid various errors. There are three kinds of functions:

  • Changing a quantity that affects some part of an interface:

    • Human population density

    • The mosquito dispersal matrix

    • The time spent matrix

  • Changing initial values, except human / host population density:

  • Changing parameter values.

Modification of any features of a model, but some of these features are:

  • change_H

  • change_calK

  • change_*par

  • change_*inits

To build a model with the features you want. The goal of software design was to make it possible to do almost anything, but there were some important constraints.

Structural Parameters

Some changes affect the model’s structure. These are considered advanced options for model building, and they are dealt with in separate vignettes. Structural changes include:

  • changing the number of population strata

  • changing the number of patches

  • adding or removing aquatic habitats

  • adding vertebrate host species

  • adding vector species