diagnostic: The 'diagnostic' function

In mu2013/KGode: Kernel Based Gradient Matching for Parameter Inference in Ordinary Differential Equations

Description

This function is used to perform diagnostic procedure to compute the residual and make diagnostic plots.

Usage

diagnostic(infer_list, index, type, qq_plot)

Arguments

infer_list	a list of inference results including ode objects and inference objects.
index	the index of the ode states which the user want to do the diagnostic analysis.
type	character containing the type of inference methods. User can choose 'rkg', 'third', or 'warp'.
qq_plot	boolean variable, enable or disable the plotting function.

Details

Arguments of the 'diagnostic' function are inference list , inference type, a list of interpolations for each of the ode state from gradient matching, and . It returns a vector of the median absolute standard deviations for each ode state.

Value

return list containing :

• residual - vector containing residual.

• interp - vector containing interpolation.

Author(s)

Mu Niu [e-mail]

Examples

## Not run: 
require(mvtnorm)
set.seed(SEED);  SEED = 19537
FN_fun <- function(t, x, par_ode) {
a = par_ode[1]
b = par_ode[2]
c = par_ode[3]
as.matrix(c(c*(x[1]-x[1]^3/3 + x[2]),-1/c*(x[1]-a+b*x[2])))
}

solveOde = ode$new(sample=2,fun=FN_fun)
xinit = as.matrix(c(-1,-1))
tinterv = c(0,10)
solveOde$solve_ode(par_ode=c(0.2,0.2,3),xinit,tinterv)

n_o = max(dim(solveOde$y_ode))
noise = 0.01 
y_no = t(solveOde$y_ode)+rmvnorm(n_o,c(0,0),noise*diag(2))
t_no = solveOde$t

odem = ode$new(fun=FN_fun,grfun=NULL,t=t_no,ode_par=rep(c(0.1),3),y_ode=t(y_no))
ktype = 'rbf'
rkgres = rkg(odem,y_no,ktype)
rkgdiag = diagnostic( rkgres,1,'rkg',qq_plot=FALSE )

## End(Not run)

mu2013/KGode documentation built on June 23, 2020, 8:04 p.m.