Evaluate change in objective function and likelihood components for up to 2 parameters.
Arguments
- fitted
MARSassess object returned by
fit_MARS()
- p1
Character string that represents the first parameter to be profiled, including the parameter name and index of the vector/array. See "Parameters" section of
make_parameters()
. Additionally, this function allows users to specifyR0_s
andh_s
(in normal units).- v1
Vector of values corresponding to
p1
- p2
Character string that represents the optional second parameter to be profiled
- v2
Vector of values corresponding to
p2
- cores
Integer for the number of cores to use for parallel processing (snowfall package)
- ...
Other argument to the base graphics function, i.e., either plot() or contour()
- x
Output from
profile.MARSassess()
- component
Character for the column in
x
to be plotted- rel
Logical, whether the relative change in
component
is plotted (TRUE) or the raw values (FALSE)- xlab
Optional character for the x-axis label
- ylab
Optional character for the y-axis label
- main
Optional character for the plot title
- plot2d
Character, plotting function for two-dimensional profiling (either a
contour()
orfilled.contour()
plot)
Value
The profile generic returns a data frame of the likelihood values that correspond to
fixed values of p1
and p2
.
Likelihood
loglike
refers to maximizing the probability of the observed data (higher values for better fit)Prior
logprior
refers to maximizing the probability of a parameter to their prior distribution (higher values are closer to the prior mode)Penalty
penalty
are values added to the objective function when parameters exceed model bounds (lower values are better)fn
is the objective function returned by RTMB (lower values are better)objective
is the objective function returned by the optimizer (lower values are better)
The accompanying plot function returns a line plot for a 1-dimensional profile or a contour plot for a two dimensional profile. Will plot the negative log likelihood or negative log prior (better fit with lower values).
Relative values are obtained by subtracting from the fitted value. See attr(x, "fitted")