Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.

Little data challenges

983 views

Published on

Little data & big data are siblings. Many of the challenges associated with big data also apply to little data however big data can provide a set of solutions for little data experiments. Effective pilot experiments are important in any evidence-based endeavor. Larger sample sizes are generally assumed to be the only solution or typically the most important. However, I propose here that there are at least three classes of solutions to little data challenges from small experiments.

Published in: Data & Analytics
  • Be the first to comment

  • Be the first to like this

Little data challenges

  1. 1. little data challenges @cjlortie
  2. 2. too much running kills
  3. 3. Too much jogging may be as bad for you as not running at all, study suggests. The Independent March 19, 2015 The (Supposed) Dangers of RunningToo Much What the data says, and what it doesn’t. Runner’s World Feb 3, 2015 Sedentary: 413 / 128
 Light: 576 / 7 Moderate: 262 / 8 Strenuous: 40 / 2
  4. 4. little data are not necessarily simple deep but not wide
  5. 5. little data challenges contrast
  6. 6. little data challenges representativeness
  7. 7. little data challenges power
  8. 8. contrast pre-post
 effect sizes to small groups to related data landscape solutions
  9. 9. contrast solutions
  10. 10. contrast solutions
  11. 11. contrast solutions
  12. 12. representativeness solutions
  13. 13. power solutions
  14. 14. power solutions online calculators to explore expectations & pilot design
  15. 15. power solutions little data studies, i.e. pilot experiments, should deep & narrow
  16. 16. dot plots by Meaghan Nolan easiest approach, increase sample size however, large samples do not replace effective designs
  17. 17. little data design implications appropriate contrasts & framing of problem Big Data & population-level contrasts use power-thinking: rejection strengths & effect sizes
  18. 18. little data can support imagination
  19. 19. can it be done must also be connected to frequency however
  20. 20. resources http://www.independent.co.uk/life-style/health-and-families/health-news/too-much-jogging-may-be-as-bad- for-you-as-not-running-at-all-study-suggests-10020478.html http://www.runnersworld.com/health/the-supposed-dangers-of-running-too-much http://hdexplore.calit2.net/wp/project/personal-data-for-the-public-good-report/ http://www.statisticalsolutions.net/pss_calc.php https://whatthestats.wordpress.com/2012/09/10/statistical-power/ https://onlinecourses.science.psu.edu/stat414/book/export/html/245 Cohen 1992. Statistical Power Analysis & A Power Primer.

×