Word of the day: Heteroscedasticity

“Data science*” is a newish field, or a new name for an old field, or maybe an umbrella name used by those of us who don’t fully grasp the bunch of tasks and jobs underneath.

A data scientist might have an IT background, or a statistics background. Generally speaking data scientists use computers to work with huge data sets and algorithms.

(Gold star for world-class oversimplification!)

Love, love, love these emerging disciplines. Spots for where corporate world’s traditional silos and boundaries have to crack and allow new cooperation. Content strategy is another area, and enterprise risk management, and on a smaller scale DevOps.

All so very interesting. Chaos is interesting when new orders emerge.

Anyhoo I am reading a ton about analytics and algorithms and data science. Very challenging.

Heteroscedasticity** describes a non-linear relationship between an Independent Variable and a Dependent Variable. This guy on the wonderfully named Stats Make Me Cry offers a nice, simple theoretical example: Age of Teenagers and Income.

Income is the Dependent Variable here.

When you change the age from 13 to 14 (a unit of 1 year), the amount of change in income variability isn’t very radical. The highest-earning 14yo doesn’t make millions more than the lowest. But if you look further down the line, changing the age from 18 to 19 (still a unit of 1 year), the range of the DV changes much more. Some 19yos are making a TON more than others.

So a 1-year change in age has a variable (different, “hetero”) effect on the change in income, depending on where you are along the axis.

Ah, his explanation is better and more clear. I’m just repeating/rewording it to help me remember the idea.


p..s. Science!

p.p.s. I must persuade my daughter to use this word in a poem.