sensPlot: Plot results of sensitivity analysis
In dkneis/mcu: Model calibration utilities

Description Usage Arguments Value Note Author(s) Examples

Plots the results of a sensitivity analysis carried out using sens.

1	sensPlot(sout, by = c("fun", "par"), nc = 1)

`sout`	A list of data frames returned from a call to `sens`.
`by`	Character string, either 'fun' or 'par'. If 'fun', the method creates one plot for each component of the objective function. If `sens` was called with a scalar-valued objective function (and thus `sout` is of length 1) only a single plot is created. If `by` is set to 'par', one plot is created for every parameter whose influence was tested. This makes sense for vector-valued objective functions only, i.e. if `sout` is of length > 1.
`nc`	The number of columns used to arrange multiple plots. This info is passed to the `layout` function.

NULL.

If by='fun', the plots can be interpreted as follows:

The value of the objective function called with default values for all parameters is plotted as a dot.
The value of the objective function called with a particular parameter at the lower limit is plotted as a blue whisker line having its origin in the dot.
The value of the objective function called with a particular parameter at the upper limit is plotted as a red whisker line having its origin in the dot.

If by='par', the plots can be interpreted as follows:

The normalized value of the objective function called with default values for all parameters is 1.
The normalized value of the objective function called with a particular parameter at the lower limit is plotted as a blue whisker line (having its origin at x=1).
The normalized value of the objective function called with a particular parameter at the upper limit is plotted as a red whisker line (having its origin at x=1).

David Kneis david.kneis@tu-dresden.de

# Sensitivity of parameters of a linear model
obs= data.frame(x=c(1,2), y=c(1,2))
model= function(p, x) { p["slope"] * x + p["intercept"] }
objfun= function(p, obs) { c(sse= sum((obs$y - model(p, obs$x))^2),
  mae= sum(abs(obs$y - model(p, obs$x)))) }
p= data.frame(
  name=c("slope","intercept"),
  default= c(1, 0.1),
  min= c(0.5, -1),
  max= c(2, 1)
)
s= sens(fn=objfun, p=p, obs=obs)
sensPlot(s, by="fun")
sensPlot(s, by="par")