run_analysis.R

# run_analysis.R
library(dplyr)
# http://archive.ics.uci.edu/ml/datasets/Human+Activity+Recognition+Using+Smartphones 

# Here are the data for the project: 
  
#  https://d396qusza40orc.cloudfront.net/getdata%2Fprojectfiles%2FUCI%20HAR%20Dataset.zip 

# You should create one R script called run_analysis.R that does the following. 

# Obtain the data, unless we already have it
# Grab our current WD, and save it; then go to the chosen work directory so 
# that we'll leave the WD where it started
oldwd <- getwd()
if (!file.exists("rawdata.zip")) {
  setInternet2(TRUE)
  sourceurl <- "https://d396qusza40orc.cloudfront.net/getdata%2Fprojectfiles%2FUCI%20HAR%20Dataset.zip"
  destfile <-"./rawdata.zip"
  download.file(url = sourceurl, destfile = destfile, method = "internal", mode = "wb")
} else {
  print("File already exists in directory, not downloading.")
}

# Unzip the rawdata.zip file, if necessary
if (!file.exists("UCI HAR Dataset")) {
print("Unzipping the raw data.")
unzip(destfile)
} else {
  print ("Dataset already unzipped.")
}
# Merges the training and the test sets to create one data set.
  # 1) Get the activity labels and features
    print("Getting the training dataset.")
    setwd("UCI HAR Dataset")
    activities <- read.table("activity_labels.txt")
    activities <- as.vector(activities$V2)
    features <- read.table("features.txt")
    features <- as.vector(features$V2)

  # 2) Read the test dataset into data
    print("Get the test dataset.")
    setwd("test")
    # testd is a data.frame holding the test data with activity numbers in first col.
    data <- read.table("X_test.txt")
    acts <- read.table("y_test.txt")
    subjects <- read.table("subject_test.txt")
    testd <- cbind(subjects,acts,data)

  # 3) Read the training dataset into traind; remove data and acts when done
    setwd("../train")
    data <- read.table("X_train.txt")
    acts <- read.table ("y_train.txt")
    subjects <- read.table("subject_train.txt")
    traind <- cbind(subjects,acts,data)

  # 4) Combine the two datasets 
    data <- rbind(testd, traind)
    rm(testd, traind)
  #    Add column names
    vars <- c("Subject_ID","Activity", as.vector(features))
    colnames(data) <- vars
# Extract the columns for means and standard deviations
  ms <- grep("mean()",names(data))
  ss <- grep("std()", names(data))
  keeps <- c(1,2,ms, ss)
  rm(ms,ss)
  data <- data[,keeps]

  #    Change first column to descriptive activity names
    print ("Adding descriptive names.")
    for(i in 1:nrow(data)){
      index <- as.integer(data[i,2])
      data[i,2] <- activities[index]
    }

# Creates a second, independent tidy data set with the average of each variable for each activity and each subject.   
  print("Creating data table frame from data frame.")
  data <- tbl_df(data)
  # Filter out each group, drop the first two columns (which we won't need once
  # we filter) and then use colMeans to get the averages.  Build this into
  # a new data frame called result

  print("Filtering and averaging.")
  result <- data.frame(row.names = colnames(data))
  for (subj in 1:30) {
    for (act in activities) {
      tempdf <- filter(data, Subject_ID == subj, Activity == act)
      tempdf <- tempdf[-c(1,2)]
      result <- rbind(result, colMeans(tempdf))
    }
  }
  # Add descriptive naes
  i <- 1
  acts <- ""
  subs <- ""
  for (subj in 1:30) { 
    for (act in activities) {
      # build vectors that we can cbind onto result
      acts <- c(acts, act)
      subs <- c(subs, subj)
    }
  }
  result <- cbind(acts[2:length(acts)], result)
  result <- cbind(subs[2:length(subs)], result)
  colnames(result) <- names(data)
  rownames(result) <- NULL

# Write the final dataset in our starting directory
print ("writing final dataset.")
setwd(oldwd)
write.table(result, file = "tidy_avg.txt", row.names = FALSE)