Free Style Data Exploration

Stages of Data Exploration - “you are given data set… what’s next”

1. Poking around - “kicking the tires” - preliminary plots and queries
2. Anything interesting? Finding candidate hypothesis
3. Evaluating the hypothesis - Statistics (coming next week)
4. Telling a story with data - presenting the results

Professor Moody Data Set as Example

What are potential questions?

1. About Professor Moody: How is he grading? How to pass the class? is this fair grading?
2. About his class - Behaviors: texting, asking questions, dozing off, participating, leaving early, coming late, etc. General analysis

Poking Around

1. Plot!
2. What plots can you make on Moody data set? And what would these plot tell you about Moody data set?

• Barplot of grades, ask_questions, on_smartphone, etc.

How many plots can one make for Professor Moody set?

Moody[grade, score, texting, questions, participation] 2 Numerical and 3 Categorical variables

1. 5 Single attribute plots

• 3 Bar plots - on Grade, Texting, and Questions
• 2 Box plots - on Score and on Participation

• Scatter plot of grade vs. participation = 1
• (3 choose 2) mosaic plots = 3
• 6 multi-box plots boxplot(numerical ~ categorical)

But much more! 65 more plots

So far, 15 plots… but now lets subset the moody data frame

1. Grade has 5 values and texting and questions have 4 values each (never, rarely, often, always)
2. How many box plots can we create for subsets of moody by (Attribute == value)? moody[moody$texting == 'always']$grade
3. 4 x 2 + 4 x 2 + 5 x 2 = 26 bar plots over subsets of moody
4. 2 x (4 + 4 + 5) = 26 box plots over subsets of moody
5. Scatter plots (grade vs. participation) on simple subsets of moody data frame (attribute = value) = 4 + 4 + 5 over subsets of moody = 13 scatter plots

So far 80 plots … just from 5 attributes

• But wait … there’s so much more!
• How about conjunctions? Subsets like “always texting and never asking questions A students”?
• This will be HUNDREDS OF PLOTS