MCestimLFSM: Numerical properties of statistical estimators operating on...

In rlfsm: Simulations and Statistical Inference for Linear Fractional Stable Motions

Numerical properties of statistical estimators operating on the linear fractional stable motion.

Description

The function is useful, for instance, when one needs to compute standard deviation of \widehat α_{high} estimator given a fixed set of parameters.

Usage

MCestimLFSM(Nmc, s, m, M, alpha, H, sigma, fr, Inference, ...)

Arguments

Nmc	Number of Monte Carlo repetitions
s	sequence of path lengths
m	discretization. A number of points between two nearby motion points
M	truncation parameter. A number of points at which the integral representing the definition of lfsm is calculated. So, after M points back we consider the rest of the integral to be 0.
alpha	self-similarity parameter of alpha stable random motion.
H	Hurst parameter
sigma	Scale parameter of lfsm
fr	frequency. Either "H" or "L"
Inference	statistical function to apply to sample paths
...	parameters to pass to Inference

Details

MCestimLFSM performs Monte-Carlo experiments to compute parameters according to procedure Inference. More specifically, for each element of s it generates Nmc lfsm sample paths with length equal to s[i], performs the statistical inference on each, obtaining the estimates, and then returns their different statistics. It is vital that the estimator returns a list of named parameters (one or several of 'sigma', 'alpha' and 'H'). MCestimLFSM uses the names to lookup the true parameter value and compute its bias.

For sample path generation MCestimLFSM uses a light-weight version of path, path_fast. In order to be applied, function Inference must accept argument 'path' as a sample path.

Value

It returns a list containing the following components:

data	a data frame, values of the estimates depending on path length s
data_nor	a data frame, normalized values of the estimates depending on path length s
means, biases, sds	data frames: means, biases and standard deviations of the estimators depending on s
Inference	a function used to obtain estimates
alpha, H, sigma	the parameters for which MCestimLFSM performs path generation
freq	frequency, either 'L' for low- or 'H' for high frequency

Examples

#### Set of global parameters ####
m<-25; M<-60
p<-.4; p_prime<-.2; k<-2
t1<-1; t2<-2
NmonteC<-5e1
S<-c(1e2,3e2)
alpha<-1.8; H<-0.8; sigma<-0.3


# How to plot empirical density

theor_3_1_H_clt<-MCestimLFSM(s=S,fr='H',Nmc=NmonteC,
                     m=m,M=M,alpha=alpha,H=H,
                     sigma=sigma,ContinEstim,
                     t1=t1,t2=t2,p=p,k=k)
l_plot<-Plot_dens(par_vec=c('sigma','alpha','H'),
                  MC_data=theor_3_1_H_clt, Nnorm=1e7)



# For MCestimLFSM() it is vital that the estimator returns a list of named parameters

H_hat_f <- function(p,k,path) {hh<-H_hat(p,k,path); list(H=hh)}
theor_3_1_H_clt<-MCestimLFSM(s=S,fr='H',Nmc=NmonteC,
                     m=m,M=M,alpha=alpha,H=H,
                     sigma=sigma,H_hat_f,
                     p=p,k=k)


# The estimator can return one, two or three of the parameters.

est_1 <- function(path) list(H=1)
theor_3_1_H_clt<-MCestimLFSM(s=S,fr='H',Nmc=NmonteC,
                     m=m,M=M,alpha=alpha,H=H,
                     sigma=sigma,est_1)

est_2 <- function(path) list(H=0.8, alpha=1.5)
theor_3_1_H_clt<-MCestimLFSM(s=S,fr='H',Nmc=NmonteC,
                     m=m,M=M,alpha=alpha,H=H,
                     sigma=sigma,est_2)

est_3 <- function(path) list(sigma=5, H=0.8, alpha=1.5)
theor_3_1_H_clt<-MCestimLFSM(s=S,fr='H',Nmc=NmonteC,
                     m=m,M=M,alpha=alpha,H=H,
                     sigma=sigma,est_3)

rlfsm documentation built on Aug. 27, 2022, 5:06 p.m.