Francis T. O'Donovan

Francis T. O'Donovan

Data Science Manager at Hospital IQ – Planet discoverer, researcher, developer, geek.

(TIL) Stats: Better descriptive statistics

less than 1 minute read

  • Instead of the mean, use the median and/or the mode.
  • Instead of the standard deviation, use the mean absolute deviation, the median absolute deviation, or the interquartile range.
  • Instead of z-scores, use percentile ranks.
  • Instead of skewness, use a QQ-plot or a histogram.
  • Instead of x standard deviations from the mean, use x median absolute deviations from the median.
  • Instead of correlation metrics, just use a scatterplot.

Via Unlearning descriptive statistics.

Comments

You May Also Enjoy

Entropy Explained, With Sheep

less than 1 minute read

All Physics students learn the Second Law of Thermodynamics, that entropy always increases. But not all such students understand why this is. To do this, let...

Machine Learning for Product Managers

1 minute read

In a previous post, I discussed the importance of learning how to properly communicate Data Science to maximize the impact of your work. Product Managers are...

Communicating Data Science with impact

less than 1 minute read

One of the major differentiators between a new Data Scientist and a more experienced one is how the more senior practitioner spends a lot of time understandi...