Heteroscedasticity
/ˌhɛtərəskɪdæsˈtɪsɪti/noun
The condition in which the spread of a variable's errors changes across the range of the data, breaking an assumption of ordinary regression.
The Greek roots
Literally: “different scattering”
The story of the word
Karl Pearson built the pair heteroscedastic and homoscedastic in 1905 from σκεδάννυμι, to scatter, the verb Homer uses for a crowd dispersing or wind driving mist apart. The spelling is a small permanent quarrel: purists write it with a k, journals mostly print the c, and nobody has ever won. Income data is the standard illustration, since rich households vary far more than poor ones.
- How it travelled
- Ancient Greek → scientific coinage → English
- First recorded
- 1905
Same family
In a sentence
The residual plot fanned out to the right, a clear sign of heteroscedasticity.