cboettig

## Error reporting
tweetError <- function(){
	system(paste('hpc-autotweets "@cboettig runtime error"', sep=""))
}
options(error = tweetError)

## Reporting a finished run

##social reporting
gitcom <- system('git commit -a -m "autocommit"', intern=TRUE)[[1]]

cboettig

#!/bin/bash
# Convert all pdf to png
for f in *.pdf
do
convert "$f" "${f%.pdf}.png"
done

cboettig

[flickr-gallery mode="search" tags="adaptivedynamics" min_upload_date="2011-01-06 00:00:00" max_upload_date="2011-01-06 23:59:59"]

cboettig

setOldClass("myS3Class")

setGeneric("foo", function(x, ...) standardGeneric("foo"))

foo =
  function(x, ...)
    UseMethod("foo")

foo.myS3Class =

cboettig

##########################################################
# King Markov, island dictator example
# this is a simple metropolis algorithm
num.visits <- 20000
population <- 1:10
current.island <- sample( 1:10 , size=1 )
visits <- {}
par(mfrow=c(2,1))
for ( i in 1:num.visits ) {
    visits[i] <- current.island

cboettig

##########################################################
# show updating process iteratively
# just paste all of this code into R to execute it

# generate data
n <- 20
y <- rnorm( n , mean=7 , sd=0.5 )
# assign prior and compute posterior as we add each y value to observations
prior.mu <- 3
k.sigma<-sd(y)

cboettig

# @file: abc.R
# @author: Carl Boettiger <cboettig@gmail.com>
# @date: 11 April, 2011
# Description: A trivial example of Approximate Bayesian Computing
# We will simulate under a Gaussian process to determine 

# The target data:
TrueMean <- 7     ## irrelevant note: "True" is rather a frequentist name for this
TrueSD <- 1
Npts <- 100

cboettig

## The observed data. 
X <- rnorm(1000, 2, 5)

loglik <- function(mu, sigma, X){
  ## The likelihood function
  sum( dnorm(X, mean=mu, sd=sigma, log=TRUE) )
}

## initial starting point 
pars <- c(mu=5, sigma=15)

cboettig

# id <- "10255/dryad.23" 
download_data <- function(id, curl=getCurlHandle() ){
# Returns the url to data file
  mets_metadata <- sprintf("%s/%s/%s", "http://datadryad.org/metadata/handle/", id, "/mets.xml")
  tt <- getURLContent(mets_metadata, curl=curl)
  page <- xmlParse(tt)
  out <- xpathApply(page, "//mets:FLocat", 
             function(x){
               link <- xmlAttrs(x, "xlink:href")

cboettig

beta <- function(i, Delta_T=1){
  1/(1+Delta_T*(i-1))
}

step_fn <- function(pars, stepsizes = .02){
# Sequential random updating 
  j <- sample(1:length(pars), 1)
  pars[j] <- rnorm(1, pars[j], stepsizes)
  pars
}