Sample sizes and stats test
Message posted by lawrence on November 3, 1999 at 12:00 AM (ET)

Hi,

I think you've got a neat site that is so helpful to all of us. Thanks.
Maybe you have answered this one but I do not seem to locate the files. I have always been troubled by the
minimum required sample size for statistical tests before the test results can be considered to be
valid. For example, I've been told that for t-tests, you need a minimum of 30 per group. Is that so?
Also is there a site on your webpage that gives the minimum size for all the different stats test?
I am particularly interested in the minimum sizes for t-tests, one-way, ANOVA (with interaction),
chi-square and Mann-Whitney.
I am working on an academic research. I cannot seem to find an answer in all the stats textbook that I have
referred to.
Thanks a million.
Lawrence

READERS RESPOND:
(In chronological order. Most recent at the bottom.)

Re: Sample sizes and stats test
Message posted by jg on November 3, 1999 at 12:00 AM (ET)

To illustrate these ideas think of the problem of estimating how much a hotel room costs in New York.
You would first have to specify what you mean - the average paid by tourists, the average cost of a specific type of room in a specific type of neighborhood, etc.

Assume that you somehow solved the above problem.
Then if you had five numbers such as 35, 47, 52, 38, and 63 you would probably feel that a reasonable average can be calculated.
However, if you had the following five number : 247, 26, 500, 45, 2000 you would probably conclude that you are comparing apples and melons and that an average
would be meaningless for these numbers.
Until you define your problem in a meaningful way simple collecting a large quantity of data is meaningless and simple confuse the issue.

Re: Sample sizes and stats test
Message posted by j-bamdad on November 5, 1999 at 12:00 AM (ET)

Re: Sample sizes and stats test
Message posted by j-bamdad on November 5, 1999 at 12:00 AM (ET)