Friday, June 12, 2009

Skedaddling

I made two discombobulating discoveries this week

1. Google produces different results for a simple one-word search, depending on the network/machine being used

2. Homo/hetero-skedacity (or -scedacity) SHOULD be homo/hetero- skedasticity (or -scedasticity)

I feel mortified though my original blog mistake was a joke about a word of which few people had heard (The 4th (1982) edition of Kendall’s Dictionary of Statistical Terms called it a “little used word")

Even when I subsequently discovered that Robert F. Engle had won the Nobel Prize in 2003 for methods of analyzing economic time series with time-varying volatility autoregressive conditional heteroskedasticity (ARCH) I failed to spot the difference. The brain often does see what it expects to see


I am not alone in this mistake. For the record, as of yesterday, on this network, Google produced the following number of hits (in thousands):

Heteroskedacity……..1.2……...Homoskedacity.........0.1 Heteroscedacity……..1.3……...Homoscedacity.........0.8
Heteroskedasticity...245.0……..Homoskedasticity...20.1
Heteroscedasticity...198.0……..Homoscedasticity...67.4





For students of my generation homoskedasticity was rather theoretical – just an assumption that was made in order for the math to work. Not something that figured much in practice, with the aids to calculation then available. You assumed it, & carried on

Heteroskedasticity became much more important & troublesome (along with autocorrelation, multicollinearity ....) for econometricians & their economic models. In other applications we never used the term (though we did often ask ‘What do the residuals look like?’ after plots became a routine part of the output of most computer packages

The notes for a time series course I went on in the late 1970s do actually contain the word heteroscedacity (sic), but this course only confirmed my desire to steer clear of any job which involved beating my head against such problems

The elision is probably easily explained by linguists. One can see (hear) how it happens. And after all, one reason why statisticians don’t own up at parties is that the very name of our subject is a terrible tongue-twisting trap, even for the perfectly sober

K or C I leave to those who care about such things, though it tempts me into another anecdote

Our Latin teacher liked to give us a bit of fun at the end of term; one year he produced a set of philology textbooks from the 1920s which had been lying still in the back of the stockroom. The fun came with the author’s passionate objection to the new-fangled abomination of the CINEMA; anybody with a classical education would know that it should be KINEMA

Oh – and I owe John Humphrys an apology

Links
Kinema
Kinema Research and Software

Welcome to Kinema Club