This package performs nonparametric regression on univariate Poisson or Gaussian data using multi-scale methods. For the Poisson case, the data x is a vector, with x_j \sim Poi(μ_j) where the mean vector μ is to be estimated. For the Gaussian case, the data x are a vector with x_j \sim N(μ_j, σ^2_j). Where the mean vector μ and variance vector σ^2 are to be estimated. The primary assumption is that μ is spatially structured, so μ_j - μ_{j+1} will often be small (that is, roughly, μ is smooth). Also σ is spatially structured in the Gaussian case (or, optionally, σ is constant, not depending on j).
The function smash
provides a minimal
interface to perform simple smoothing. It is actually a wrapper to
smash.gaus
and smash.poiss
which
provide more options for advanced use. The only required input is
a vector of length 2^J for some integer J. Other options include
the possibility of returning the posterior variances, specifying a
wavelet basis (default is Haar, which performs well in general due
to the fact that smash uses the translation-invariant transform)
Matthew Stephens and Zhengrong Xing
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.