Get Data Project Assignment

Author: Mário O. de Menezes

This project asked us to get a bunch of data files (data sets), process them and yield a tidy data set, as defined during the course lectures.

The dataset used for this project represent data collected from the accelerometers from the Samsung Galaxy S smartphone. It was downloaded from the Project description page, and its link is:

https://d396qusza40orc.cloudfront.net/getdata%2Fprojectfiles%2FUCI%20HAR%20Dataset.zip

After extracting the zip file, a directory structure is created, and the base folder name is UCI HAR Dataset.

The run_analysis.R script needs the to be inside this folder, together with the features.txt file.

To run the script, source it inside a R session. It might be necessary to use setwd() function in order to source command be able to find the run_analysis.R script, or give source the full pathname to it.

The script works as follow:

1. First, all files are read (X_test, X_train, y_test, y_train, sub_test, sub_train, features)
2. Some duplicated names there exist in the features file (this file contains the variable names - 561 names); they're removed to ensure that select function works correctly. These variable names will not be used, since they are related to bandsEnergy (not mean or std).
3. After removing the duplicated columns, the remaining names are assigned to the columns of the X_testU dataset (X_testU and X_trainU are the cleaned dataset, without duplicated columns). sub_test and sub_train data sets have also column names assigned accordingly.
4. There're now 6 datasets ready for merging to yield 2 complete datasets that will be latter combined. Using the cbind function, the TestDF and TrainDF dataframes are formed, keeping the columns on both in the same order.
5. According to the experiment description, the test dataset contains 30% of the subjects and the train dataset, 70%. Inspecting both dataframes, we confirmed that there's no overlapping between them, so we need just to combine them on a row basis. This is done using the rbind function, yielding one more dataframe: MergeDF.
6. As required by the Project assignment, only columns containing mean and std measurements should be used in the tidy dataset. So, we used select to filter out only columns that contains mean firstly (MeanDF) and then std in sequence (StdDF). Here's the reason we de-duplicated column names in an early step.
7. These two new dataframes were then merged (cbind) and the PersonId and Activity columns were added back to them.
8. As required, all activities were correctly labed using the activity_labels.txt file. This was accomplished using the recode function from the car package.
9. The dataframe was then grouped by PersonId and Activity variables, and the summarise_each function was used to summarize the data, computing the mean value of each column for each subject and activity.
10. The resulting dataset is a dataframe with 180 observations and 88 variables, with PersonId and Activity as the first two columns. This dataset is then written to the "step5tidy.txt" file using the write.table function.