This class represents the linear trend model X_t=bt + mu + e_t, t=1,...,n, with intercept mu~N(mu.mu0,mu.sd^2) and slope b~N(b.mu1,b.sd^2) and e_t being a nuisance parameter time series.
nuisanceModel_linearTrend(mu.mu0 = 0, mu.sd = 10000, b.mu1 = 0,
b.sd = 1, prop.scaling = 1)A priori mean and standard deviation of mu
A priori mean and standard deviation of b
Scaling parameter for generating Metropolis-Hastings proposals of parameter of interest theta=c(mu,b)
S3 nuisanceModel object representing the model parameter theta=c(mu,b)
of interest, containing the following fields:
Dimension of parameter of interest (here: theta_dim=2)
Logical; Should the outermost Fourier frequencies be
ignored in the frequency domain representation? (here: excludeBoundary=F)
Function taking the two arguments data,theta
to compute the nuisance/noise time series e_t from data and parameter
theta of interest. (here: e_t=data-mu-b*(1:n))
Function taking the parameters data
(Numeric vector of input data), f (Numeric Vector of current
spectral density at the Fourier frequencies within the Gibbs sampling algorithm)
and previous_theta (Previously sampled value of c(mu,b)) and
returning a new proposal value for c(mu,b)
Function taking the Numeric Vector data of input
data as argument to generate an initial value for c(mu,b) to start an MCMC
algorithm
Function; Log density of prior of theta
The parameters mu and b are assumed to be a priori independent.
The returned object of this function is intended for usage within
gibbs_AR_nuisance, gibbs_NP_nuisance
and gibbs_NPC_nuisance.
The method propose_next_theta is optimized to be close to the
marginal joint posterior of (mu,b) in the model.
The proposal scaling can be controlled with the parameter prop.scaling,
where larger values yield a broader (smaller values yield narrower)
proposal distribution.