mrecos

!image

mrecos

require(caret)

#load some data
data(USArrests)

### Prepare Data (postive observations)
# add a column to be the strata.  In this case it is states, it can be sites, or other locations
# the original data has 50 rows, so this adds a state label to 10 consecutive observations
USArrests$state <- c(rep(c("PA","MD","DE","NY","NJ"), each = 5))
# this replaces the existing rownames (states) with a simple numerical index

mrecos

# I started to play with the data and used this to get it a bit cleaned up and prepared for analysis
# I am sure my code is not the most efficient or prettiest, but if there are errors, please let me know
# Perhaps you will find this usefull in saving a few minutes getting the data ready for the fun part; analysis!
# DATA : https://www.opendataphilly.org/dataset/parking-violation

library(data.table)
library(lubridate)
setwd(<<YOUR DIRECTORY>>) # Set WD if desired
data <- fread("Parking_Violations.csv")
dat <- data.frame(data) # convert to a data.frame from data table

mrecos

# Here is my attempt to use ggplot2 to recreate Randy Olson's (@randal_olson), recreation of the famous Tufte New York City Weather chart.
# The original Tufte chart is here:
# http://www.edwardtufte.com/bboard/q-and-a-fetch-msg?msg_id=00014g
# Randy Olson recreated this chart for a post on 538 (http://fivethirtyeight.com/) linked here:
# http://fivethirtyeight.com/features/what-12-months-of-record-setting-temperatures-looks-like-across-the-u-s/?ex_cid=538twitter
# and Randy posted his data and Python code here:
# https://github.com/mrecos/data/tree/master/us-weather-history
# This is a ggplot implementation of Randy's basic design, but without the top and right side axis and legend
# I think the legend in Randy's plot may be added in Illustrator of something
# The only real hard-code (aside from .csv location) here is the y-axis labels that show the data is from July '14 to June '15, 

mrecos

# The original plot in question appears to be the results of a model
# based on scores and age, a distribution is predicted for score on
# each age class.  The range of predicted scores is visualized as a
# normal'ish density over each age class.  There is also a point which is the 
# mean observation of the raw score at each age class.  This does not need to
# be at the mean of the distribution depending on fit of the model. 
# There is also a line that is the predicted response of model over age classes.
# 
# I did not have the data or model from this example, so I improvised.  
# I first created sythetic model data based on the example, and from the 

mrecos

load.fun <- function(x) { 
  x <- as.character(x) 
  if(isTRUE(x %in% .packages(all.available=TRUE))) { 
    eval(parse(text=paste("require(", x, ")", sep=""))) 
	print(paste(c(x, " : already installed; requiring"), collapse=''))
  } else { 
	if(x == "twitteR"){
		print(paste(c(x, " : not installed; installing"), collapse=''))
		install_github("twitteR", username="geoffjentry")
		print(paste(c(x, " : installed and requiring"), collapse=''))

mrecos

# Driving distance from multuple starts to multilpe destinations
# modified from the code of gmapsdistance package  by Rodrigo Azuero Melo
# https://cran.r-project.org/web/packages/gmapsdistance

gdist2 <- function (origin, destination, mode, key) 
{
  typeKEY = class(key)
  if (typeKEY != "character") {
    stop("Key should be string")
  }

mrecos

lut_match <- function(LUT, code_field, desc_col = 2, is_NA = "NA", no_code = "N/A"){
require(stringr)
require(splitstackshape)
  # this loops through all unique codes in code_field.  Potentially slow
  code_field[which(code_field == "")] <- NA
  unique_code <- unique(code_field)
  coded <- code_field
  for(i in seq_along(unique_code)){
    if(is.na(unique_code[i])){
      coded[is.na(coded)] <- is_NA # set this to whatever works

mrecos

# function to generate samples of data
get_data <- function(n, exp_i, intercept, beta1, sigma, plot = TRUE){
  library(lattice)
  x <- sample(seq(1,100,0.1), n, replace = TRUE)
  gen_model <- intercept + beta1 * x^(exp_i)
  y <- rnorm(n, gen_model, sigma)
  dat <- data.frame(x, y)
  if(isTRUE(plot)){
    print(xyplot(y~x, type=c("smooth", "p"),col.line = "darkorange", lwd = 2))
  }

mrecos

library('foreach')
library('ggplot2')
library('gganimate')
# library('animation') # require for original code
library('reshape2')
library('doParallel') # for multicore support

## Guts and working concept from post below.  
## http://johnramey.net/blog/2011/06/05/conways-game-of-life-in-r-with-ggplot2-and-animation/#comments
## I updated ggplot, melt, varnames, gganimate, and faded cells