runtsir: runtsir

In tsiR: An Implementation of the TSIR Model

Description

This function runs the TSIR model.

Usage

runtsir(
  data,
  xreg = "cumcases",
  IP = 2,
  nsim = 10,
  regtype = "gaussian",
  sigmamax = 3,
  userYhat = numeric(),
  alpha = NULL,
  sbar = NULL,
  family = "gaussian",
  link = "identity",
  method = "deterministic",
  inits.fit = FALSE,
  epidemics = "cont",
  pred = "forward",
  threshold = 1,
  seasonality = "standard",
  add.noise.sd = 0,
  mul.noise.sd = 0,
  printon = F,
  fit = NULL,
  fittype = NULL
)

Arguments

data	The data frame containing cases and interpolated births and populations.
xreg	The x-axis for the regression. Options are 'cumcases' and 'cumbirths'. Defaults to 'cumcases'.
IP	The infectious period in weeks. Defaults to 2 weeks.
nsim	The number of simulations to do. Defaults to 100.
regtype	The type of regression used in susceptible reconstruction. Options are 'gaussian', 'lm' (linear model), 'spline' (smooth.spline with 2.5 degrees freedom), 'lowess' (with f = 2/3, iter = 1), 'loess' (degree 1), and 'user' which is just a user inputed vector. Defaults to 'gaussian' and if that fails then defaults to loess.
sigmamax	The inverse kernal width for the gaussian regression. Default is 3. Smaller, stochastic outbreaks tend to need a lower sigma.
userYhat	The inputed regression vector if regtype='user'. Defaults to NULL.
alpha	The mixing parameter. Defaults to NULL, i.e. the function estimates alpha.
sbar	The mean number of susceptibles. Defaults to NULL, i.e. the function estimates sbar.
family	The family in the GLM regression. One can use any of the GLM ones, but the options are essentially 'poisson' (with link='log'), 'gaussian' (with link='log' or 'identity'), or 'quasipoisson' (with link='log'). Default is 'gaussian'.
link	The link function used with the glm family. Options are link='log' or 'identity'. Default is 'identity'.
method	The type of next step prediction used. Options are 'negbin' for negative binomial, 'pois' for poisson distribution, and 'deterministic'. Defaults to 'deterministic'.
inits.fit	Whether or not to fit initial conditions using simple least squares as well. Defaults to FALSE. This parameter is more necessary in more chaotic locations.
epidemics	The type of data splitting. Options are 'cont' which doesn't split the data up at all, and 'break' which breaks the epidemics up if there are a lot of zeros. Defaults to 'cont'.
pred	The type of prediction used. Options are 'forward' and 'step-ahead'. Defaults to 'forward'.
threshold	The cut off for a new epidemic if epidemics = 'break'. Defaults to 1.
seasonality	The type of contact to use. Options are standard for 52/IP point contact or schoolterm for just a two point on off contact, or none for a single contact parameter. Defaults to standard.
add.noise.sd	The sd for additive noise, defaults to zero.
mul.noise.sd	The sd for multiplicative noise, defaults to zero.
printon	Whether to show diagnostic prints or not, defaults to FALSE.
fit	Now removed but gives a warning.
fittype	Now removed but gives a warning.

Examples

require(kernlab)
London <- twentymeas[["London"]]
## Not run: 
plotdata(London)
res <- runtsir(data=London,method='pois',nsim=10, IP=2,inits.fit=FALSE)
plotres(res)

## End(Not run)

Example output

Loading required package: kernlab
           alpha        mean beta         mean rho         mean sus 
        9.70e-01         1.08e-05         4.57e-01         1.14e+05 
prop. init. sus. prop. init. inf. 
        3.01e-02         6.12e-05 
Warning: Ignoring unknown aesthetics: fill

tsiR documentation built on Jan. 21, 2021, 1:06 a.m.