This script solves the course project Analysis and Data Cleaning Coursera.

Data were taken from the test and training of human physical activity records on a group of people.

Data were obtained from information provided by accelerometers in Galaxy S smartphones.

IMPORTANT: this script uses the libraries:

reshape2

stringr

This script reads the files:

features.txt

X_train.txt

X_test

activity_labels.txt

y_train.txt

y_test.txt

subject_train.txt

subject_test.txt

Located within the folder: UCI Dataset HAR

features.txt presents the name of 561 features defined in this study.

X_train.txt presents the results of measurements of 561 features for each subject in training.

X_test.txt presents the results of measurements of 561 features for each subject in testing.

activity_labels.txt describes the names or labels of 6 activities developed by each subject in this experiment.

y_train.txt describes the names or labels of the activities of each of the subjects in training

y_test.txt describes the names or labels of the activities of each of the subjects in testing

subject_train.txt identifies the subject developing each training activity

subject_test.txt identifies the subject developing each testing activity

The script creates datasets for training and testing and extracts only the measures for mean

and standard deviations for each measure of the original data

Using gsub functions, the names of the activities are reassigned and the names of the features are redefined.

The new names of the features are more descriptive than the original.

Example: the old name of the feature: "tBodyAcc-mean()-X" was

redefined as : "Timedomain.BodyAcceleration.Average.X"

An independent tidy data set with the average of each feature of each activity is created and named

"means_ttotal.txt". This file is created when the script is run as the Samsumg data is in the working directory

Read Samsumg data

file_features <- "UCI HAR Dataset/features.txt"
features <- read.table(file_features, header = F, sep=" ",colClasses="character")
file_train <- "UCI HAR Dataset/train/X_train.txt"
file_test <- "UCI HAR Dataset/test/X_test.txt"
file_activity_labels <- "UCI HAR Dataset/activity_labels.txt"
X_train <- read.table(file_train, header=F, dec=".", colClasses="character")
X_train <- as.data.frame(sapply(X_train, as.numeric))
X_test <- read.table(file_test, header=F, dec=".", colClasses="character")
X_test <- as.data.frame(sapply(X_test, as.numeric))
activity_labels <- read.table(file_activity_labels, header=F, colClasses="character")

Assign features names

colnames(X_train) <- features$V2
colnames(X_test) <- features$V2

file_y_train <- "UCI HAR Dataset/train/y_train.txt"
yy_train <- read.table(file_y_train, header=F, colClasses="character")
file_y_test <- "UCI HAR Dataset/test/y_test.txt"
yy_test <- read.table(file_y_test, header=F, colClasses="character")
file_subject_test <- "UCI HAR Dataset/test/subject_test.txt"
subject_test <- read.table(file_subject_test, header=F, colClasses="character")
file_subject_train <- "UCI HAR Dataset/train/subject_train.txt"
subject_train <- read.table(file_subject_train, header=F, colClasses="character")

Assign activity labels and subjects for eaach activity

X_train$activity <- factor(yy_train$V1, labels= activity_labels$V2)
X_train$subject <- subject_train$V1

X_test$activity <- factor(yy_test$V1, labels= activity_labels$V2)
X_test$subject <- subject_test$V1

library(reshape2)

Consolidate test and train data for subject and activity with melt

dataMelt_test <- melt(X_test, id=c("subject", "activity"), measure.vars = features$V2)
dataMelt_train <- melt(X_train, id=c("subject", "activity"), measure.vars = features$V2)

calculate mean of train and test data for each combination of activity and for each subject using dcast function

pryData_train_mean <- dcast(dataMelt_train, activity_subject ~ variable, mean)
pryData_test_mean <- dcast(dataMelt_test, activity_subject ~ variable, mean) # calcula para cada combinacion de activity y subject el valor del promedio de la variable seleccionada antes

Select the "mean" columns and rename the labels with more readable names in training data

df <- pryData_train_mean
library(stringr)
df1 <- df[,grep("mean", colnames(df))]
colnames(df1) <- gsub(pattern="tBody", replacement="Timedomain.Body", colnames(df1))
colnames(df1) <- gsub(pattern="tGravi", replacement="Timedomain.Gravi", colnames(df1))
colnames(df1) <- gsub(pattern="fBody", replacement="Frequencydomain.Body", colnames(df1))
colnames(df1) <- gsub(pattern="-", replacement=".", colnames(df1))
colnames(df1) <- gsub(pattern="mean", replacement="Average", colnames(df1))
colnames(df1) <- gsub(pattern="Acc", replacement="Acceleration", colnames(df1))
colnames(df1) <- str_replace_all(colnames(df1), "([()])","")

Select the "std" columns and rename the labels with more readable names in training data

df2 <- df[,grep("std", colnames(df))]
colnames(df2) <- gsub(pattern="tBody", replacement="Timedomain.Body", colnames(df2))
colnames(df2) <- gsub(pattern="tGravi", replacement="Timedomain.Gravi", colnames(df2))
colnames(df2) <- gsub(pattern="fBody", replacement="Frequencydomain.Body", colnames(df2))
colnames(df2) <- gsub(pattern="-", replacement=".", colnames(df2))
colnames(df2) <- gsub(pattern="mean", replacement="Average", colnames(df2))
colnames(df2) <- gsub(pattern="Acc", replacement="Acceleration", colnames(df2))
colnames(df2) <- str_replace_all(colnames(df2), "([()])","")

df_train_mean <-cbind(df1,df2)

NA columns in data frame are eliminated

df_train_mean <- df_train_mean[,apply(df_train_mean,2, function(x){!all(is.na(x))})]

df_train_mean$subject <- pryData_train_mean$subject
df_train_mean$activity <- pryData_train_mean$activity

df <- pryData_test_mean

Select the "mean" columns and rename the labels with more readable names in testing data

df1 <- df[,grep("mean", colnames(df))]
colnames(df1) <- gsub(pattern="tBody", replacement="Timedomain.Body", colnames(df1))
colnames(df1) <- gsub(pattern="tGravi", replacement="Timedomain.Gravi", colnames(df1))
colnames(df1) <- gsub(pattern="fBody", replacement="Frequencydomain.Body", colnames(df1))
colnames(df1) <- gsub(pattern="-", replacement=".", colnames(df1))
colnames(df1) <- gsub(pattern="mean", replacement="Average", colnames(df1))
colnames(df1) <- gsub(pattern="Acc", replacement="Acceleration", colnames(df1))
colnames(df1) <- str_replace_all(colnames(df1), "([()])","")

Select the "std" columns and rename the labels with more readable names in testing data

df2 <- df[,grep("std", colnames(df))]
colnames(df2) <- gsub(pattern="tBody", replacement="Timedomain.Body", colnames(df2))
colnames(df2) <- gsub(pattern="tGravi", replacement="Timedomain.Gravi", colnames(df2))
colnames(df2) <- gsub(pattern="fBody", replacement="Frequencydomain.Body", colnames(df2))
colnames(df2) <- gsub(pattern="-", replacement=".", colnames(df2))
colnames(df2) <- gsub(pattern="mean", replacement="Average", colnames(df2))
colnames(df2) <- gsub(pattern="Acc", replacement="Acceleration", colnames(df2))
colnames(df2) <- str_replace_all(colnames(df2), "([()])","")

df_test_mean <-cbind(df1,df2)

NA columns in data frame are eliminated

df_test_mean <- df_test_mean[,apply(df_test_mean,2, function(x){!all(is.na(x))})]

df_test_mean$subject <- pryData_test_mean$subject
df_test_mean$activity <- pryData_test_mean$activity

Some test values as references

df_train_mean[df_train_mean$subject=="25" & df_train_mean$activity=="WALKING_UPSTAIRS","Frequencydomain.BodyAcceleration.std.Y"]

df_train_mean[df_train_mean$subject=="25" & df_train_mean$activity=="WALKING","Frequencydomain.BodyAcceleration.Average.Y"]

df_train_mean[df_train_mean$subject=="22" & df_train_mean$activity=="WALKING","Frequencydomain.BodyAcceleration.std.X"]

df_train_mean[df_train_mean$subject=="1" & df_train_mean$activity=="WALKING","Timedomain.BodyAcceleration.Average.X"]

Creates the independent tidy data set with average of each of each feature(variable) for each activity and

for each subject consolidating training and testing data

Labels names are redefined in the consolidated Merged data with the extensions .Train and . Test

for training and testing features

The file "means_ttotal.txt" in the main directory stores the tidy data set with the averages of each variable

for each activiyy and each subject

mergedData <- merge(df_test_mean, df_train_mean, by.x=c("subject", "activity"), by.y=c("subject", "activity"), all =T) #pega
colnames(mergedData) <- gsub(pattern=".x", replacement=".Train", colnames(mergedData))
colnames(mergedData) <- gsub(pattern=".y", replacement=".Test", fixed=TRUE, colnames(mergedData))
write.table(mergedData, "means_ttotal.txt")