Wufoo

Name * First Last
Data

The data for this exam comes from the Open Data Visualization hosted by Mozilla (you will probably not need additional information on the data beyond what is given in the exam, but feel free to visit this website: http://design-challenge.mozillalabs.com/open-data/OpenDataCompetition.php to learn more about the competition / data )

The data is organized in three different files, named events.csv, users.csv, and survey.csv.
All three files have a variable 'user_id' (or 'id'), that refer to the same set of users.

The data can be found at http://www.hofroe.net/mozilla
Question #1 -
events.csv is a file of more than 3 GB of data, i.e. it is not feasible to completely load it into R (don't try it!).

Instead, write an iterative approach to read in the data 10000 lines at a time until you have read in the complete data for the first 100 users (events.csv is sorted by user_id)

It is safe to assume that the data for 100 users fits into main memory.

Hint:
you might find the command rbind very useful, see ?rbind
Question #2 - 1

use the commands
filename <- "http://www.hofroe.net/mozilla/events.csv"
res <- scan(filename, what=character(), sep="\n", nlines=100)
to read in the first 100 lines of the file events.csv

Describe the structure of res:
Question #2 - 2

Using res, come up with a dataset 'df' with the following variables:
- the same column(s) as res
- separate all values of each line into variables, i.e. variables 'user_id', 'event.code', 'data1', ..., 'data4', 'timestamp' and 'session_id'
Make sure that all variables of the resulting dataset are of the appropriate data type.

Hint: functions ldply and/or unlist might come in handy
Question #2 - 3

Some elements in data1 end in the text '.dll'. Introduce a new binary variable DLL in the dataset df that is TRUE, if data1 ends in '.dll', and FALSE otherwise. Compare DLL to event.code.
Question #2 - 4

Some records of the dataset df show the string "". Assume that this corresponds to a missing value.

Introduce a new variable NMIS in the dataset df, that contains the number of non-missing elements in each record.
Question #3 - 1

out of the events for the first 100 users select only events for users who are also in the survey (use events100.csv, if you could not get the code for the first question to run).

Extend the data with information on each of the users from users.csv and survey.csv
Question #3 - 2

From the extended file, select data for five users at random. Set the code up in a way that (with a high probability) you will get five different users, if you re-run the code.
Plot event_code by timestamp, connect all events from the same session (session_id), colour by operating system and facet by user.
Attach your commented (!) R script containing all the commands for the above questions
Do Not Fill This Out

Wufoo

Data