interv_fit: Interpolating the process, given an interval averaged series

In c-foschi/mlintervals: Interpolation of interval averaged data using a maximum likelihood method

Description

Maximum Likelihood interpolation of a Wiener process, given a series of observed integrals over adiacent time intervals of equal lenght. To get point estimates of the interpolating function, use method predict. The method assumes a process with homegeneous variance, but the estimated interpolating line is quite robust to heteroskedasticity, anyway, also scale parameter sigma2 is estimated as well as SEs based on it.

Usage

interv_fit(M, t_0 = 0, t_unit = 1, t_end = NULL)

Arguments

M	series of observed integrals
t_0	time at beginning of the first interval
t_unit	time span of each interval
t_end	time at the end of the last interval. If t_end is specified, t_unit is ignored

Details

predict

Value

A list of class interv_fit, with the following attributes:

• $data: series M

• $knots: estimated value for the Wiener process in the points between intervals

• $sigma2: estiamated scale parameter

• $se: standard errors for the knots values. They depend on sigma2 and on homoskedasticity assumption

• $covariances: covariances of consecutive knots estimates.

Examples

M= c(0, 1, -1, 2, 4)
mod= interv_fit(M)
plot(mod)

mod= interv_fit(M, t_0= 5, t_unit= 2)
plot(mod)

c-foschi/mlintervals documentation built on July 1, 2020, 1:25 a.m.