willtownes

# Simulation of a Paintball Game.
# The two players start off in random positions. They move one square each turn. Then they shoot at each other.
# Probability of who shoots first is 50/50. 
# Probability of a hit is proportional to the distance between the players.
# Tracks history of last 3 moves.
# Prints out graphs to show the movement of the players and the shots.
# Game over when someone is hit.

shoot<-function(a.pos,b.pos){
  # returns TRUE if the shot was a hit, otherwise FALSE

willtownes

My background

I am interested in biological problems that require both computational and statistical methods. I started out doing tropical ecology. Then I was a software tester for a while, where I learned programming. I took a bunch of statistics classes at night (two of my favorites were Bayesian Statistics and Stochastic Simulation). I am a fan of open source collaboration and plan on using software engineering techniques like automated testing, version control and continuous integration in my future research. Here are some things I want to learn more about:

Statistics

• Bayesian hierarchical modeling and probabilistic graphical models
• Bayesian nonparametrics, including all the stochastic processes theory beneath it (Dirichlet processes, etc)
• Nonparametric and semiparametric statistics in general
• Variational Inference- I find this an intriguing alternative to MCMC
• Spatial Statistics (I love looking at maps)

• How does the choice of projection affect things? Is there a way to do 3-D spatial

willtownes

#sigma2, df, and Cp may not be reliable in default elasticnet function enet
#provide replacements
#S2L is the estimate for sigma2 based on maximal model (full_ols_model from lm())
S2L<- sum(residuals(full_ols_model)^2)/full_ols_model$df.residual

Cp_enet<-function(enet_obj,y,X,S2L){
  #returns the Mallows Cp statistics for a given elastic net object
  betamat<-enet_obj$beta.pure
  #create a matrix where each column is prediction at one of the df levels
  yhats<-X%*%t(betamat) + mean(y)

willtownes

#!/bin/bash
#SBATCH -n 1
#SBATCH -t 400
#SBATCH -p serial_requeue
#SBATCH --mem=35000
#SBATCH --mail-type=END,FAIL
#SBATCH --mail-user=my_username@mail.harvard.edu

#take filename from command line argument
FILENAME=$1

willtownes

#!/bin/bash

LOGFILE="master_job_ids.log"

echo "======Starting new analysis run========" >> $LOGFILE
names=( SRR1592185 SRR1592187 SRR1592189 SRR1592186 SRR1592188 SRR1592190 )
for ((i=0; i<${#names[@]} ; i++)) do
	echo "Running Batch Job for ${names[i]}. SLURM Job ID shown below:" >> $LOGFILE
	#make results directory
	mkdir -p ${names[i]}_results

willtownes

"""
Python Script to extract code blocks from tex files for CS281
input: a file in current working directory with suffix *.tex
output: the same file with suffix *.py, with everything except the python parts removed
"""
#import pdb
from sys import argv
import os

def convert(ifile,ofile):

willtownes

#Example of using recursion to apply newton's method to an arbitrary function in order to find the zeros of the function. If no zeros are found after some number of iterations, NA is returned

newton.recursive<-function(func,deriv,counter=100,x=1,relchange=1){
    diff = func(x)/deriv(x)
    relchange = abs(diff/x)
    if(counter==0){
      print("Failed to converge")
      return(NA)
    }
    else if(relchange <= 1e-10){

willtownes

'''I recently had to deploy a test script I had written in python 2.7 to a server where only python 2.4 was allowed. In the process of making my code backwards compatible, I made a brief checklist of things to watch out for:
* any and all functions can still be used but you have to manually code them
* with-as blocks must be replaced with try-finally
* try-except-finally must be converted to try-except-else
* string formatting: "".format() must be replaced with "%s"%(args) approach.
* Replace all instances of "except Exception as err:" with "except Exception, e:"
* may need to manually convert some unicode strings into str, especially when loading json
* Some modules have different names, eg: "email.MIMEText" versus "email.mime.text.MIMEText"
* The datetime module does not have strptime() in 2.4. However, you can import it with a different name and then make a subclass with the desired behavior by calling strptime() from the time module instead:'''

willtownes

index2slag<-function(index){
  # Takes an index between 1 and 32 and converts it into a binary vector. This is used to iterate all of the possible values of (s_t,s_t1,s_t2,s_t3,s_t4)
  num<-paste(rev(intToBits(index-1)),"")
  return(as.logical(as.numeric(num[28:32])))
}

ref.s<-t(sapply(1:32,index2slag)) #table of all S_t1,... combinations
ref.s #structure of this reference table is crucial in step5!

slag2index<-function(slag){

willtownes

### Adaptive Rejection Sampling
# by Will Townes

rexp_trunc<-function(n,slope=-1,lo=0,hi=Inf){
  #draw n samples from the truncated exponential distribution
  #the distribution is proportional to exp(slope*x)
  #default is standard exponential
  #slope cannot equal zero
  #lo is lower truncation point, can be -Inf
  #hi is upper truncation point, can be +Inf