Tables for analysing data

Question 1

What are the characteristics of two-way-tables?

Answer 1

Organizes two categorical variables.
• Row variable: across (→)
• Column variable: down
Shows counts (= absolute frequencies) or percent (=relative frequencies)

Question 2

What is a joint distribution?

Answer 2

• Distribution for each cell

* Divide number in cell by total sample size

Question 3

What is a marginal distribution?

Answer 3

• Distribution for a single variable

* Divide total of a variable by total sample size

Question 4

What is a conditional distribution?

Answer 4

• Relative frequencies

* E.g. given that the person is a boy what is the probability his eyes are not green? (8/10=0.8)

Question 5

Two-Way tables: no association?

Answer 5

• Knowledge about x does not give any information about probability of y
• Joint distribution can be calculated by multiplying marginal distributions and dividing the product by the table total

Question 6

Two-way tables: perfect association?

Answer 6

• By knowing x you can perfectly predict y
• Fields have to be 0 for perfect association to
occur

Question 7

What is the special case of interrupter reliability? (Kappa)

Answer 7

• Reliability: To what extend a repeated measurement yields the same score.
• 2 raters judge a same set of objects independently
• They agree or disagree on different objects and you want to find out how much they agree (=interrater reliability =agreement Kappa)

Question 8

What is the agreement Kappa (k)?

Answer 8

If the raters judged the objects randomly, there is still chance of agreement = expected agreement.
Kappa takes out the random factor via subtracting this expected random chance
𝑘 = (𝐴𝑜𝑏𝑠𝑒𝑟𝑣𝑒𝑑 − 𝐴𝑒𝑥𝑝𝑒𝑐𝑡𝑒𝑑 ) /(𝑁 − 𝐴𝑒𝑥𝑝𝑒𝑐𝑡𝑒𝑑)

Question 9

How do you calculate Kappa?

Answer 9

1. N= Total sum of objects (100) 2. 𝑨𝒐𝒃𝒔𝒆𝒓𝒗𝒆𝒅:
   • Add up all diagonal cells (9+14+30)
2. 𝑨𝒆𝒙𝒑𝒆𝒄𝒕𝒆𝒅:
   • Multiply the two marginals of each Variable
   and divide by N.
   (i.e.: (26x24)/100)
   • Add up all results. (ie. 6.24+11.84+16.38)
3. Plug the results in formula