NetNews Usenet Archive 1993 #3/NN_1993_3.iso/spool/bit/listserv/statl/2439

Internet Message Format | 1993-01-25 | 1.8 KB

Path: sparky!uunet!pipex!unipalm!uknet!comlab.ox.ac.uk!ph
From: [e-mail] (Patrick Haggard)
Newsgroups: bit.listserv.stat-l
Subject: Violated t-test assumptions: how to adjust?
Message-ID: <[e-mail]>
Date: 25 Jan 93 19:55:06 GMT
Sender: [e-mail] (Patrick Haggard)
Organization: Physiology Department, Oxford University, Oxford, UK.
Lines: 27
Originator: [e-mail]

In his classic 1960 paper, Boneau discussed the effects on the t test
and related tests of violating the parametric assumptions.  From the
inferential point of view, I remember the most drastic problem is the
situation of unequal sample sizes, where the smaller sample has the
larger variance.  This leads to increased type I error: in his
examples from 0.06 to 0.16, as I recall.

Well this is exactly the situation in which I now find myself.  I
have a sample of 5 and an unrelated sample of 20, and the former has
the greater variance.  So I thought I would just ask what the current
state of opinion is on this situation.  In particular, can anyone
suggest an adjustment to the t statistic which takes this problem into
account, and will give a more accurate probability value.

I originally read Boneau in my undergraduate days: are there any more
up to date papers on violating parametric assumptions.

Many thanks
Patrick
--------------------------------------------------------------------------
Patrick Haggard                         Email (WORLD): [e-mail]
University Laboratory of Physiology     Email (JANET): [e-mail]
Parks Road,                             Tel. [phone]
Oxford, OX1 3PT                         Fax. [phone]
England
--------------------------------------------------------------------------