tomclegg.net


Diary
Examples
Hire Tom
Mostly Mozart
    70watts
    alphabet
    batteries
    binary
    bluesky
    conservation
    darwin
    decibel
    doppler
    e
    educatiolution
    entropy
    eureka
    fallout
    hanoi
    infinitesimal
    iq
    kcrtech
    lightyearsahead
    littlec
    logic1
    logic2
    lpcd
    mass
    normal
    npcomplete
    optics
    periodictable
    pope
    psa
    reaction
    refrigerators
    resistance
    revealedreligion
  >scientificsurvey<
    senses
    spacetime
    spam
    specialrelativity
    stars
    statistics1
    string
    uncertainty
    unlikely
    warning
    water
    waves
Patches
School
Scrapbook
Software
Telephones




colocation
comments
davidireland
edsgranola
faq
funsites
goodlooking
goodmovies
google-earth-saucy-amd64
houserules
liberating
resume
resume2
scratch
shopping
snacks
todo
university
warisbogus

Scientific Survey
Posted November 18, 2001

I read something on the back of a newspaper the other day that claimed to be the result of a "scientific survey" about a new recreation facility. Presumably the word "scientific" was thrown in just to make it look important.

There are several interesting pieces of information missing from the report. The first thing I looked for was a confidence figure: something along the lines of, "this survey has a 99% chance to be 98% accurate." You can get that out of a formula, and it's supposed to give you an idea of how big the random sample was and stuff like that.

If you don't have a confidence figure, you can at least look at the size of the sample and make a subjective judgement about whether it was appropriate. But this particular report said nothing about the sample size. Did they give a questionnaire to 71 people? Or maybe 1000? Or 17? Who knows.

Truly random sampling of a human population is practically impossible. The people who don't have telephones are rarely surveyed, for example; or the people who prefer not to respond to questionnaires. The best you can do without a lot of extra trouble is keep track of how many people refused to answer the questionnaire, or didn't answer the phone at all. That would give a rough idea of how random the sample really was. Of course, if you don't know the sample size, the number of refusals doesn't mean much to you. So that's probably why that figure wasn't mentioned either.

Of course, the most important thing missing from the article was the questionnaire itself.

So perhaps the piece should have been labelled, "City Hall's interpretation of a scientific survey, whose actual results are available at City Hall to anyone who's interested."

So I'm off to City Hall to get a copy of the questionnaire. I'll leave you with...