Coursera Getting and Cleaning Data Project Submission
How to run
Simply load and run the run_analysis.R script. Output will be produced in the same directory, called tidydata.txt
Step by Step Commentary
Initialisation
The script first of all downloads and unzips the data again. Although the spec says it's ok to assume the samsung data is in the same directory, this was a simple way to know exactly where in the directory structure the script is running
Step 1 - Merge Test and Training sets
This step first column-binds each features dataset to its corresponding activity ID and subject ID columns to produce full test and training sets, then row-binds the test and training sets together to produce a full dataset called dat.
Step 2 - extract only mean and standard deviation measurements.
This step finds the indices for the calculated means and standard deviations from the original dataset. meanFreq() items are not included because they appear to have come from a meanFreq() function which indicates some other type of data. I made this choice because now all means have a corresponding standard deviation in the final dataset.
Step 3 - name the activities within the dataset
This step merges the activity names from act_labels with the main dataset, using the activity IDs to join on. The first column of the dataset produced is the activity ID column used to join on, which is then discarded as it has been replaced by a factor column with the activity names in.
Step 4 - label the variables
This step first gets the original variable names using the indices from step 2, appending subject and activity as headers for those columns. It then uses string replacement to go through the variable names and make them conform with the requirements of given in the first lecture for next week on the final page of [these notes] (https://d396qusza40orc.cloudfront.net/getdata/lecture_slides/04_01_editingTextVariables.pdf)
I had originally planned to stick with the original names but there's been quite a lot of discussion on the forums for instance in [this thread] (https://class.coursera.org/getdata-008/forum/thread?thread_id=177) which suggest the TAs think it's best to demonstrate you can update the variable names to something that is to some eyes more human readable eg expanding abbreviations.
A lot of reading around the forums did suggest that this was the approach favored by the TAs for this particular assignment, although personally I would hope to discuss with my data clients whether camelCase naming would be more appropriate.
Step 5 - summarize
Again in the forums there's been quite a bit of discussion about whether tidy data means thin and long or wide and short, the general consensus seems to be that as long as it is one or the other (ie not a mixed up mess) it's ok. I have chosen wide and short in this case because it reduces the number of repetitions of the identifier fields, this makes the dataset take up a bit less space but can make using some of the plyr manipulation functions a little trickier.