View pulse.Rproj
Version: 1.0
RestoreWorkspace: No
SaveWorkspace: No
AlwaysSaveHistory: No
EnableCodeIndexing: Yes
UseSpacesForTab: Yes
NumSpacesForTab: 2
Encoding: UTF-8
View gp_regression.stan
functions{
// GP: computes noiseless Gaussian Process
vector GP(real volatility, real amplitude, vector normal01, int n_x, real[] x ) {
matrix[n_x,n_x] cov_mat ;
real amplitude_sq_plus_jitter ;
amplitude_sq_plus_jitter = amplitude^2 + 1e-6 ;
cov_mat = cov_exp_quad(x, amplitude, 1/volatility) ;
for(i in 1:n_x){
cov_mat[i,i] = amplitude_sq_plus_jitter ;
}
View read_incrementally.R
fname = 'temp.txt'
#wait for it to exist
while(!file.exists(fname)){}
old = 0
new = old
while(T){
while(new==old){
new = file.size(fname)
}
f = file(
View find_last_line.R
f = file(description='bigFile.txt',open='rb')
n = 0
temp = 1
while(length(temp)<2){
n = n + 1
seek(con=f,origin='end',where=-(2^n))
temp <- scan(f,what='character',quiet=T)
}
last_line = temp[2]
View gp_regression.stan
functions{
# GP: computes noiseless Gaussian Process given pre-computed unique distances
vector GP(real volatility, real amplitude, vector normal01, int n_x, int n_dx, vector dx_unique, int [,] dx_index) {
# covars: unique entries in covariance matrix
vector[n_dx] covars ;
# covMat: covariance matrix
matrix[n_x,n_x] covMat ;
View crf_demo.R
library(tidyverse)
library(rstan)
rstan_options(auto_write = TRUE)
curve(dweibull(x,shape=1.75,scale=1e3),from=0,to=3e3)
one_response = dweibull(x=1:3e3,shape=1.75,scale=1e3)
one_response = one_response/max(one_response)
# create a signal with pulses at t=2e3, t=4e3 & t=5e3
obs = rep(0,8e3)
View bigStan.R
# Code below may not be up to date, see ezStan (https://github.com/mike-lawrence/ezStan/blob/master/R/bigStan.R) for latest version
# todo:
# during-sampling: effective sample size and rhat for each parameter
# during-sampling: diagnostics?
#usage:
# #compile the model using rstan::stan_model:
View cfa.stan
data{
# n_subj: number of subjects
int n_subj ;
# n_y: number of outcomes
int n_y ;
# y: matrix of outcomes
matrix[n_subj,n_y] y ;
# n_fac: number of latent factors
int n_fac ;
# y_fac: list of which factor is associated with each outcome
View compare_ceq_dx_udx.R
#load packages
library(tidyverse)
library(rstan)
rstan_options(auto_write = TRUE)
# Make some fake data ----
#set random seed for reproducibility
set.seed(1)
View gp_example.R
#load packages
library("curl")
library("ggplot2")
library("rstan")
#retrieve the data
tmpf = tempfile()
curl_download("http://www.metoffice.gov.uk/hadobs/hadcrut4/data/current/time_series/HadCRUT.4.5.0.0.annual_ns_avg.txt", tmpf)
gtemp = read.table(tmpf, colClasses = rep("numeric", 12))[, 1:2] # only want some of the variables
names(gtemp) = c("Year", "Temperature")