# Association Analysis

CA3 was to test the comprehension of association analysis. The CA had two components to it:

1. Complete 15 SWIRL ‘R’ Programming questions
2. Complete 5 questions on the concepts of Lift, Chi Squared and other association analysis algorithms.

1. SWIRL ‘R’ Programming questions

What is SWIRL?

Swirl is a software package for the R programming language that turns the R console into an interactive learning environment. Users are provided with real-time feedback as they are guided through self-paced practicals in the fields of data science and R programming.

Who is swirl aimed at?

Swirl is aimed at beginners in R programming.

What’s needed to use swirl?

You will need a computer, an Internet connection and a recent version of R installed on your machine and you’re good to go.

If you are interested in trying R programming using swirl, see details here

1. Five questions on the concepts of Lift, Chi Squared and other association analysis algorithms.

Q1. LIFT Analysis

Please calculate the following lift values for the table correlating Burger & Chips below:

• LIFT (Burger, Chips)
• LIFT (Burger, ^Chips)
• LIFT (^Burger, Chips)
• LIFT (^Burger, ^Chips)

 Column1 Chips ^Chips Total Row Burgers 600 400 1000 ^Burgers 200 200 400 Total Column 800 600 1400

LIFT (Burgers, Chips)

s (Burgers u Chips) = 600/1400 = 0.428

s(Burgers) = 1000/1400 = 0.714

s(Chips) = 800/1400  = 0.571

LIFT (Burgers, Chips) = 0.428/(0.714*0.571) = 1.049

LIFT (Burgers, Chips) > 1

My answer suggests that Burgers and Chips are positively correlated.

LIFT (Burgers, ^Chips)

s(Burgers u ^Chips) = 400/1400 = 0.285

s(Burgers) = 1000/1400 = 0.714

s(^Chips) = 600/1400 = 0.428

LIFT (Burgers, ^Chips) = 0.285/(0.714*0.428) = 0.932

LIFT (Burgers, ^Chips) < 1

My answer suggests that Burgers and ^Chips are negatively correlated.

LIFT (^Burgers, Chips)

s(^Burgers u Chips) = 200/1400 = 0.142

s(^Burgers) = 400/1400 = 0.285

s(Chips) = 800/1400 = 0.571

LIFT (^Burgers, Chips) = 0.142/(0.285*0.571) = 0.872

LIFT (^Burgers, Chips) < 1

My answer suggests that ^Burgers and Chips are negatively correlated.

LIFT (^Burgers, ^Chips)

s(^Burgers u ^Chips) = 200/1400 = 0.142

s(^Burgers) = 400/1400 = 0.285

s(^Chips) = 600/1400 = 0.428

LIFT (^Burgers, ^Chips) = 0.142/(0.285*0.428) = 1.164

LIFT (^Burgers, ^Chips) > 1

My answer suggests that Burgers and Chips are positively correlated.

Q2. Please calculate the following lift values for the table correlating Ketchup & Shampoo below:

• LIFT (Ketchup, Shampoo)
• LIFT (Ketchup, ^Shampoo)
• LIFT (^Ketchup, Shampoo)
• LIFT (^Ketchup, ^Shampoo)

 Column1 Shampoo ^Shampoo Total Row Ketchup 100 200 300 ^Ketchup 200 400 600 Total Column 300 600 900

LIFT (Ketchup, Shampoo)

s(Ketchup u Shampoo) = 100/900 = 0.111

s(Ketchup) = 300/900 = 0.333

s(Shampoo) = 300/900 = 0.333

LIFT (Ketchup, Shampoo) = 0.111/(0.333*0.333) = 1.001

LIFT (Ketchup, Shampoo) = 1

My answer suggests that Ketchup and Shampoo are independent.

LIFT (Ketchup, ^Shampoo)

s(Ketchup u ^Shampoo) = 200/900 = 0.222

s(Ketchup) = 300/900 = 0.333

s(^Shampoo) = 600/900 = 0.666

LIFT (Ketchup, ^Shampoo) = 0.222/(0.333*0.666) = 1.001

LIFT (Ketchup, ^Shampoo) = 1

My answer suggests that Ketchup and Shampoo are independent.

LIFT (^Ketchup, Shampoo)

s(^Ketchup u Shampoo) = 200/900 = 0.22

s(^Ketchup) = 600/900 = 0.67

s(Shampoo) = 300/900 = 0.33

LIFT (^Ketchup, Shampoo) = 0.222/(0.666*0.333) = 0.22/0.22 = 1.001

LIFT (Ketchup, Shampoo) = 1

My answer suggests that Ketchup and Shampoo are independent.

LIFT (^Ketchup, ^Shampoo)

s(^Ketchup u ^Shampoo) = 400/900 = 0.444

s(^Ketchup) = 600/900 = 0.666

s(^Shampoo) = 600/900 = 0.666

LIFT (^Ketchup, ^Shampoo) = 0.444/(0.666*0.666) = 1.001

LIFT (Ketchup, Shampoo) = 1  (Ketchup and Shampoo, Independent)

My answer suggests that Ketchup and Shampoo are independent.

Q3. Chi Squared Analysis

Please calculate the following chi Squared values for the table correlating Burger and Chips below (Expected values in brackets).

• Burgers & Chips
• Burgers & Not Chips
• Not Burgers & Chips
• Not Burgers & Not Chips

For the above options, please also indicate if each of your answer would suggest independent, positive or negative correlation.

 Column1 Chips ^Chips Total Row Burgers 900 (800) 100 (200) 1000 ^Burgers 300 (400) 200 (100) 500 Total Column 1200 300 1500

Chi-squared = ∑ (observed-expected) 2/ (expected)

Χ= (900-800)/ 800 + (100-200)/ 200 + (300-400)/ 400 + (200-100)/ 100

= 100/ 800 + (-100)/ 200 + (-100)/ 400 + 100/ 100

= 10000/800 + 10000/200 +10000/400 + 10000/100 = 12.5 + 50 + 25 + 100 = 187.5

Burgers & Chips, correlated  Χ2  > 0.

Expected 800, Observed 900, Burgers & Chips – Negatively Correlated.

Expected 200, Observed 100, Burgers & ^Chips – Negatively Correlated.

Expected 400, Observed 300, ^Burgers & Chips – Negatively Correlated.

Expected 100, Observed 200, ^Burgers & ^Chips – Negatively Correlated.

Q4: Chi Squared Analysis

Please calculate the following chi squared values for the table correlating burger and sausages below (Expected values in brackets).

• Burgers & Sausages
• Burgers & Not Sausages
• Sausages & Not Burgers
• Not Burgers and Not Sausages

For the above options, please also indicate if each of your answer would suggest independent, positive correlation, or negative correlation?

 Column1 Chips ^Chips Total Row Burgers 800 (800) 200 (200) 1000 ^Burgers 400 (400) 100 (100) 500 Total Column 1200 300 1500

Χ= (800-800)/ 800 + (200-200)/ 200 + (400-400)/ 400 + (100-100)/ 100

= 0/ 800 + 0/ 200 + 0/ 400 + 0/ 100 = 0

Burgers & Chips, Independent Χ2  = 0.

Burgers & Chips– Observed & Expected, 800 – Independent

Burgers & ^Chips – Observed & Expected, 200 – Independent

^Burgers & Chips – Observed & Expected, 400 – Independent

^Burgers & ^Chips – Observed & Expected, 100 – Independent

Q5:

Under what conditions would Lift and Chi Squared analysis prove to be a poor algorithm to evaluate correlation/dependency between two events?

The conditions under Lift & Chi Squared analysis that could prove to be a poor algorithm to evaluate correlation / dependency between two events are when there are too many null transactions observed.

Please suggest another algorithm that could be used to rectify the flaw in Lift and Chi Squared?

Another algorithm that could be used to rectify the flow in Lift & Chi squared is: AllConf, Cosine, Jaccard, MaxConf, Kulczynski.

Aside:  Regarding the gif at the beginning of this blog post, the word burgers is mentioned 55 times through out the post, so I found it’s use appropriate.

References

Swirl, (2017). Available at: http://swirlstats.com/students.html (Accessed Mar. 2017).

Royal with cheese (2014). [Online image]. Available at: http://metro.co.uk/2014/09/18/national-cheeseburger-day-7-of-the-best-movie-quotes-featuring-cheeseburgers-4871752/ (Accessed Mar. 2017).