CodeBook.md

##Course Project: clean up activity dataset

To speed up the merge of train and test dataset, first subset both dataset to the required 66 variables and then add the activtiy and subject variable to both table. Afterwards just use the join function from plyr library to do the merge.

1. Find rows of all mean() and std() measurement from feature.txt, create a table featureTable_Std_Mean . The "Nr" column matches the column position in X_train.txt and X_test.txt.

featureTable<-read.table("features.txt",header = FALSE,sep="",col.names = c("Nr","Function"))
featureTable_Std_Mean <- featureTable[featureTable$Nr %in% grep("std\\(\\)|mean\\(\\)",featureTable$Function),]

2. Use the featureTable_Std_Mean$Nr vector to subset the X_train.txt and X_test.txt with the wanted variables in table x_train_sub and x_test_sub

x_train <- read.table("X_train.txt",head=FALSE,sep="",col.names=featureTable$Function)
x_test <- read.table("X_test.txt",head=FALSE,sep="",col.names=featureTable$Function)
x_train_sub<-x_train[,featureTable_Std_Mean$Nr]
x_test_sub<-x_test[,featureTable_Std_Mean$Nr]

4. read file y_train.txt, y_test.txt, activity_labels.txt into table

y_train<-read.table("y_train.txt",head=FALSE,col.names = ("id"))
y_test<-read.table("y_test.txt",head=FALSE,col.names=("id"))
activity_labels<-read.table("activity_labels.txt",head=FALSE,sep="",col.names = c("id","activity"))

5. Add a variable activity into x_train_sub and x_test_sub

activity_train<-activity_labels[y_train$id,]
activity_test<-activity_labels[y_test$id,]
x_train_sub$activity<-activity_train$activity	
x_test_sub$activity<-activity_test$activity

6. Create 2 table subject_train and subject_test from subject_train.txt and subject_test.txt

subject_test<-read.table("subject_test.txt",head=FALSE,col.names=("subject_id"))
subject_train<-read.table("subject_train.txt",head=FALSE,col.names=("subject_id"))

7. Add a variable subject into x_train_sub and x_test_sub by using vector subject_train$subject_id and subject_test$subject_id

x_train_sub$subject<-subject_train$subject_id
x_test_sub$subject<-subject_test$subject_id

8. Merge x_train_sub and x_test_sub into new dataframe allData

library(plyr)
allData<-join(x_train_sub,x_test_sub,type="full")

9. melt allData to calcutlate average

allDataMelt<-melt(allData,id.vars = c("subject","activity"),measure.vars = c(1:66))
allDataAverage<-dcast(allDataMelt,c("activity","suject") ~ variable,mean)
write.table(allDataAverage,file = "allDataAverage.txt")