Upgrade to Pro — share decks privately, control downloads, hide ads and more …

Barista Camp 2015 | Christopher Hendon - Performing a good experiment

Barista Camp 2015 | Christopher Hendon - Performing a good experiment

Chris introduces us to the act of performing a good experiment with four clearly recommended steps. Highlighting potential failures on the way and using the audience to demonstrate how to conduct a basic statistical experiment, the viewer can feel confident that Hendon's science background is serving us well. Closing his talk with the proclamation that he has no vested interest in the business of coffee and is motivated by wanting everyone to ‘make better coffee’, we hope he sticks around!

50ebb032d4bf0548ae4298f5712b8445?s=128

Barista Guild of Europe

March 21, 2016
Tweet

Transcript

  1. Performing a good experiment Christopher H. Hendon Department of Chemistry

    Massachusetts Institute of Technology hendon@mit.edu
  2. Boston, Massachusetts, USA 84 Nobel Laureates over the years Ranked

    #1 university in the world
  3. Some MIT inventions 1.  The World Wide Web (Tim Berners-Lee)

    2.  Email (Ray Romlinson) 3.  The Transistor Radio (William Shockley) 4.  Oil Refining 5.  Gillette Razors 6.  GPS (Ivan Getting) 7.  E-ink (now found in the kindle) 8.  Human Genome Project
  4. •  Learn to ask a good question •  Determine dependent/independent

    variables •  Design and execute experiment •  Analyze data Experimental Science
  5. •  Learn to ask a good question •  Determine dependent/independent

    variables •  Design and execute experiment •  Analyze data Experimental Science
  6. A good question has many important features –  What are

    we trying to figure out? What can we currently not explain –  Why are we trying to figure this out? What can we gain from exploring this problem –  Does the question have an answer? A ‘Yes’ and ‘No’ must be distinguishable –  Do we have the ability to answer it? Cost and expertise associated with the answer Asking a good question
  7. •  Learn to ask a good question •  Determine dependent/independent

    variables •  Design and execute experiment •  Analyze data Experimental Science
  8. Some variables can be directly modulated, others are implicitly affected

    by this change. e.g. Flow rate of water through a group head vs. % mass extracted (using a refractometer) Easiest ‘trick’ to identifying variables is to simply try and draw a graph The variables
  9. None
  10. None
  11. None
  12. None
  13. None
  14. •  Learn to ask a good question •  Determine dependent/independent

    variables •  Design and execute experiment •  Analyze data Experimental Science
  15. We want our experiments to contribute towards an ‘end goal’.

    Any individual experiment does not give the full story. Think big, test small
  16. •  Learn to ask a good question •  Determine dependent/independent

    variables •  Design and execute experiment •  Analyze data Experimental Science
  17. Conclusions are only ever relative. You can ‘see what you

    want’ in the data. Data Analysis: It’s not personal
  18. The Birthday Paradox Two assumptions: All birthdays are equally probable

    Everyone tells the truth Listen to each other. The birthday paradox
  19. Outcome: Does this mean that 2 / __________ people have

    a birthday on ____________? No. Statistical sampling is tricky stuff! The birthday paradox
  20. 50% probability with 23 people 99.9% probability with 70 people

  21. You must try and fault yourself

  22. % Particle Diameter (µm) Make sure you check the fine

    details •  EK-43 grinding Las Ilusions: Volume %
  23. % Particle Diameter (µm) •  EK-43 grinding Las Ilusions: Volume

    % (dotted line is Surface area %)
  24. % Particle Diameter (µm) •  EK-43 grinding Las Ilusions: Volume

    % (dotted line is counts in %)
  25. % Particle Diameter (µm) Make sure you check the fine

    details % GOOD DATA BAD DATA
  26. Not all EK-43’s are the same (…sorry Christian)

  27. Dangerous to conclude information from data that has not been

    appraised by other specialists with no financial conflicting interests. There is no peer-review in coffee
  28. An Idea: The Coffee Experiment To elucidate the difference in

    extraction of Ethiopian versus Kenyan coffees.
  29. An Idea: The Coffee Experiment When you have BOTH Ethiopian

    and Kenyan coffee Dial in one of the coffees to whatever ratio you think tastes delicious. Then brew the other coffee with the exact same parameters and measure the extraction with a refractometer.
  30. None
  31. None
  32. None
  33. Visit my website

  34. Christopher H. Hendon

  35. Christopher H. Hendon

  36. Comments, Questions, Criticism