About the project

This is a project performed in SKKU Design and Analysis of Experiments (STA 5031).

• Conduct a real randomized experiment, collect data, and analyze them.
• Present a non-statistical paper using an (quasi-)experiment in the field outside statistics and critize them.
• Bring my own research that is related to an (quasi-)experiment.

I chose real experimental design, the first one.

Coffee-Electrocardiogram Experiment

Goal of the Experiment

• Does caffeine affect electrocardiogram (ECG) or average heart rate?
• Caffeine: ☕ drinking capsule coffee 40 ml
• ECG result: ⌚ average heart rate

Design

We implement Latin square design. For example,

• A, B, C, D once in every row,
• once in every column

About factors

• Latin square: 4 by 4
• 2 blocking factors
  • Row: Coffee (40 ml) to water ratio
    1. 1:0 (Espresso)
    2. 1:2.5 (Water 100 ml)
    3. 1:5 (Water 200 ml)
    4. 1:7.5 (Water 300 ml)
  • Column: Drinking speed
    1. <=5 sec
    2. 5-15 sec
    3. 15-30 sec
    4. 30< sec
• Interesting factor: Intake of caffeine from Starbucks by Nespresso
  1. House blend: 74.5 mg
  2. Sumatra: 54.5 mg
  3. Decaf espresso roast: 3 mg
  4. None or water: 0 mg
• Randomly allocate these treatments 1, 2, 3, and 4 to A, B, C, and D
  • assign these to above table
  • random treatment assignment

set.seed(1)
sample(LETTERS[1:4])
#> [1] "A" "C" "D" "B"

Output

• Measure ECG using Apple watch Series 4: See https://support.apple.com/en-us/HT208955
• Output
  • Average Heart rate (in BPM) difference between after and before taking the coffee
  • log return might be better (after got feedback)

Controlling other variables

• Drink coffee every morning (between 8:30 a.m. and 9:00 a.m. KST), after eating a piece of bread
• When Measuring ECG,
  • sit at my desk
  • rest arms on a my desk
• Use same strip for apple watch: Nike sport band of same fit
• Nespresso machine: Pixie C61 in my home

For more details

Click Slides on the 🔝.

• Project: Github repository
• Slides: presentation pdf
• Code: Source codes including Rmd for the slide
• Dataset: datsets by this experiment including preprocessed ones
• Appendix: Supplementary material related to the dataset