swthresh              package:SpherWave              R Documentation

_T_h_r_e_s_h_o_l_d_i_n_g _o_f _S_p_h_e_r_i_c_a_l _W_a_v_e_l_e_t _D_e_c_o_m_p_o_s_i_t_i_o_n ('_s_w_d') _O_b_j_e_c_t

_D_e_s_c_r_i_p_t_i_o_n:

     This function performs various ways to threshold a `swd' class
     object.

_U_s_a_g_e:

     swthresh(swd, policy, by.level, type, nthresh, value = 0.1, 
     Q = 0.05) 

_A_r_g_u_m_e_n_t_s:

     swd: an object of class `swd'

  policy: threshold technique. At present the possible policies are
          `"universal"', `"probability"', `"fdr"', `"Lorentz"' and
          `"sure"'.

by.level: If FALSE, then perform a global threshold. If TRUE, a
          thresholding value is computed and applied separately to each
          resolution level.

    type: the type of thresholding. This can be `"hard"', `"soft"' or
          `"Lorentz"'.

 nthresh: the number of resolution levels to be thresholded in the
          decomposition

   value: the user supplied threshold represented by quantile level for
          `"probability"' policy

       Q: parameter for the false discovery rate of `"fdr"' policy

_D_e_t_a_i_l_s:

     This function thresholds or shrinks details stored in a `swd'
     object and returns the thresholded details in a modified `swd'
     object. For level-dependent thresholding, `"universal"',
     `"Lorentz"' and `"fdr"' are provided. Only hard type thresholding
     is proper for `"probability"' thresholding. Also note that only
     soft type thresholding is proper for `"sure"' thresholding.

_V_a_l_u_e:

     An object of class `swd'. This object is a list with the following
     components.  

     obs: observations

  latlon: grid points of observation sites in degree

  netlab: vector of labels representing sub-networks

     eta: bandwidth parameters for Poisson kernel

  method: extrapolation methods, `"ls"' or `"pls"'

  approx: if TRUE, approximation is used.

grid.size: grid size (latitude, longitude) of extrapolation site

  lambda: smoothing parameter for penalized least squares method

      p0: starting level for extrapolation. Resolution levels p0+1,
          ..., L is used for extrapolation.

 gridlon: longitudes of extrapolation sites in degree

 gridlat: latitudes of extrapolation sites in degree

 nlevels: the number of multi-resolution levels

   coeff: interpolation coefficients

   field: extrapolation on grid.size

density1: density of SBF

  latlim: range of latitudes in degree

  lonlim: range of longitudes in degree

  global: List of successively smoothed data

 density: density of SW coefficients

  detail: List of details at different resolution levels

 swcoeff: spherical wavelet coefficients

thresh.info: thresholding information and ranges of local components
          before thresholding

_R_e_f_e_r_e_n_c_e_s:

     Oh, H-S. and Li, T-H. (2004) Estimation of global temperature
     fields from scattered observations by  a spherical-wavelet-based
     spatially adaptive method. _Journal of the Royal Statistical
     Society Ser._ B, *66*, 221-238.

_S_e_e _A_l_s_o:

     'sbf', 'swd', 'swr'.

_E_x_a_m_p_l_e_s:

     ### Observations of year 1967
     #data(temperature)
     #names(temperature)

     # Temperatures on 939 weather stations of year 1967    
     #temp67 <- temperature$obs[temperature$year == 1967] 
     # Locations of 939 weather stations    
     #latlon <- temperature$latlon[temperature$year == 1967, ]

     ### Network design by BUD
     #data(netlab)

     ### Bandwidth for Poisson kernel
     #eta <- c(0.961, 0.923, 0.852, 0.723, 0.506)

     ### SBF representation of the observations by pls
     #out.pls <- sbf(obs=temp67, latlon=latlon, netlab=netlab, eta=eta, 
     #    method="pls", grid.size=c(50, 100), lambda=0.89)

     ### Decomposition
     #out.dpls <- swd(out.pls)

     ### Thresholding
     #out.univ <- swthresh(out.dpls, policy="universal", by.level=TRUE, 
     #    type="hard", nthresh=4) 

