library(plyr) library(stringr)

the following functions load the neccessary datas(test and train datas) into R objects. You have to enter

the file address on your computer in the quotation marks instead of the text instructions

test <- read.table("put test dataset file path here") subject.test <- read.table("put test_subject file path here") activity.test <- read.table("put test_activity file path here") train <- read.table("put train dataset file path here") subject.train <- read.table("put train_subject file path here") activity.train <- read.table("put train_activity file path here") features <- read.table("put feature file path here")

Before anything else lets lable the test and train dataset variable according to features

names(test) <- features$V2 names(train) <- features$V2

#moreover: names(subject.test)[1] <- "subject" names(subject.train)[1] <- "subject" names(activity.test)[1] <- "activity" names(activity.train)[1] <- "activity"

first of all we have to grab the columns with the means and stds. using grepl() to filter them from the

features object and then apply that logical vector to the test.set and train.set objects to subset the

required columns

filter <- grepl("mean|std", features$V2) test <- test[, filter] train <-train[, filter]

#now we have to produce another column to act as a joining point for each test and train group datasets test$seq <- seq(along = test[,1]) subject.test$seq <- seq(along = activity.test[,1]) activity.test$seq <- seq(along = activity.test[,1]) train$seq <- seq(along = train[,1]) subject.train$seq <- seq(along = subject.train[,1]) activity.train$seq <- seq(along = activity.train[,1])

now time to merge each dataset using plyr package join_all() function

dflist.test <- list(test, subject.test, activity.test) dflist.train <- list(train, subject.train, activity.train) test.merged <- join_all(dflist.test) train.merged <- join_all(dflist.train) test.merged <- test.merged[, c(80:82, 1:79)] train.merged <- train.merged[, c(80:82, 1:79)]

now in an attemp to merge the two data sets together we have to change their sequence a little bit

test.merged$seq <- seq(7353, 10299)

time to merge

complete.data <- merge(test.merged, train.merged, all = T)

One little last step to convert activity numbers to their descreptive labels using Regular expressions

complete.data$activity <- sub(1, "WALKING", complete.data$activity) complete.data$activity <- sub("2", "WALKING_UPSTAIRS", complete.data$activity) complete.data$activity <- sub("3", "WALKING_DOWNSTAIRS", complete.data$activity) complete.data$activity <- sub("4", "SITTING", complete.data$activity) complete.data$activity <- sub("5", "STANDING", complete.data$activity) complete.data$activity <- sub("6", "LAYING", complete.data$activity)

So, finally in an attemp to run mean() function on subject-activity pairs sensor records I'm going to create

a new variable called 'pairs' to give us the factors for pair-wise summarization of data

pairs <- paste(as.character(complete.data$subject), complete.data$activity) pairs <- factor(pairs)

Then I will write a function to gather the means by pairs on every column using tapply

runner <- function (DataFrame, index) { final <- data.frame(pairs = levels(pairs)) count <- 0 for ( cols in DataFrame) { count = count + 1 keep <- tapply(cols, index, mean) keep <- as.data.frame(keep) names(keep)[1] <- names(DataFrame)[count] keep$pairs <- levels(pairs) final <- merge(final, keep, by = "pairs", all = T) } return(final) }

Let's call it!

runner(complete.data[,4:82], pairs)