[TowerTalk] My Yearly Mistake

Al Kozakiewicz akozak at hourglass.com
Sun Aug 16 16:12:52 EDT 2015


>Which is why knowing the distribution.... 

As not everything has a Gaussian (normal; bell) distribution. Much of what is taught in the typical statistics course for non-mathematicians focusses on normal distributions and their variants because they nicely represent so much of real life populations of measurements.

W/R/T median and average, there is no "right" one.  It all depends on the nature of what it is you're measuring and what you want to know.  The median is useful for a lot of sociological data as the question often asked is "what is typical for a single data point".  For populations that are additive in nature (say, rainfall amounts), the average is probably more useful if for no other reason than you can multiply it by the number of samples and get the total for the dataset.

Al
AB2ZY


More information about the TowerTalk mailing list