Newly tracked file
R/utilities.R
changed.
Other files ignored by Codecov
man/Amwg.Rd
has changed.
33  33  
34  34  #' Turn a nonadaptive Metropolis sampler into an adaptive Metropolis sampler 

35  35  #' 

36    #' Given a nonadpative sampler of the form f(..., s), \code{Amwg} will return a 

36  +  #' Given a nonadaptive sampler of the form f(..., s), \code{Amwg} will return a 

37  37  #' function g(...) that automatically adapts the Metropolis proposal standard 

38    #' deviation s to try and acheive a target acceptance rate. 

38  +  #' deviation s to try and achieve a target acceptance rate. 

39  39  #' 

40  40  #' \code{Amwg} uses the Adaptive MetropolisWithinGibbs algorithm from Roberts 

41  41  #' & Rosenthal (2009), which rescales the proposal standard deviation after a 

42  42  #' fixed number of MCMC iterations have elapsed. The goal of the algorithm is 

43    #' to acheive a target acceptance rate for the Metropolis step. After the 

43  +  #' to achieve a target acceptance rate for the Metropolis step. After the 

44  44  #' n\emph{th} batch of MCMC iterations the log of the proposal standard 

45  45  #' deviation, \eqn{log(s)}, is increased/decreased by \eqn{\delta(n)}. 

46  46  #' \eqn{log(s)} is increased by \eqn{\delta(n)} if the observed acceptance rate 
71  71  #' by \eqn{f}. This functionality can be used, for example, if \eqn{f} samples 

72  72  #' each of its returned elements individually, updating each element using a 

73  73  #' Metropolis step. See the examples for an illustration of this use case. In 

74    #' such settings, \eqn{f} should be contructed to receive \eqn{s} as a vector 

74  +  #' such settings, \eqn{f} should be constructed to receive \eqn{s} as a vector 

75  75  #' argument. 

76  76  #' 

77  77  #' @param f nonadaptive Metropolis sampler of the form f(..., s) 
Files  Coverage 

R  98.14% 
Project Totals (2 files)  98.14% 