pirate-stats

  1. Correlation doesn’t imply causation. Just because pirates decreased as world temperatures increased doesn’t mean the two are related.
  2. Random sampling is everything.
  3. 40 randomly selected individuals is enough to ballpark it. 100 is better. Much over 500 or so and you’re wasting resources without getting better information.
  4. Confidence intervals are too wonkish. Convert it to a margin of error so everyone knows what you mean.
  5. By the way, margin of error? It’s important.
  6. Make a Likert scale with an odd number of responses. That way, someone can choose the middle option that represents “no opinion”.
  7. And, Likert scale is pronounced “LICK-ert” not “LIKE-ert”.
  8. Your sample indicates the population, but is never a 100% dead-on result.
  9. Use the mean as a measure of central tendency unless you’ve got some crazy outliers. Then use the median.
Advertisements