THE SOCIAL SCIENCE PERCENTAGE NUISANCE. BROADSHEET No. 6 AUGUST i972

Geary, R. C. (1972) THE SOCIAL SCIENCE PERCENTAGE NUISANCE. BROADSHEET No. 6 AUGUST i972. UNSPECIFIED.

PDF - Published Version Download (873Kb)

Abstract

One of the episodes in a long lifetime in statistics which, on recoiiection,,gives the writer most satisfaction was a forthright attack he delivered many years ago at ~a~Dublin Rotary Club Luncheon on Opini0n-type polls. The immediate provocation was a banner headline in a newspaper "XY Poll Does It Again"m in a UK general election. The argument ran that as XY prophesied that Labour would poll 49 per cent of the votes when the actual poll turned out to be 5o per cent (both figures imaginary) this represented an error of"only .I p e,rcent ."~ It was easy to point out the falsityof this claim: no one in his senses.wouldhave anticipated Labour’s polling as much as 55 per cent. oras little as 45 per cent~ a range of io per cent (and it might be less). The ;’i per cent error" should be related to this io per cent (or less) giving a real error of io per cent (or more), seriously raising the question of whether the XY poll had any value at all.

Actions (login required)