Re: T-test and robustness
- From: daniel_stahl@xxxxxx
- Date: Wed, 26 Sep 2007 03:35:42 -0700
Have a look at this online simulation program.
http://onlinestatbook.com/stat_sim/robustness/index.html
It will convince you that normality is less an issue for the t-test
than unequal variances, independent of sample size. If the
distribution nonnormal but similar between the two groups the ttest is
pretty robust., there is almost no alpha error inflation or
reduction. See Howell (1997) Statistical Methods for Psychologists
for a summary and recommendations.
Best wishes, Daniel
cat.. wrote:
Dear colleagues,
I remember I read, in the past, somewhere, that the student test was
quite robust regarding violations of the normality assumption but
cannot remember where I read this. So it is still relevant to use it
even with 'small' sample sizes (not too small however, maybe 15 or 20
in groups to be compared).
Does one has an opinion on the matter and/or could provide me with a
reference ?
Regards,
Catherine.
.
- References:
- T-test and robustness
- From: cat..
- T-test and robustness
- Prev by Date: Re: Converting Likert-scale items to continuous scale
- Next by Date: Download Database for multi-way tables
- Previous by thread: T-test and robustness
- Next by thread: Ordinal regression primer?
- Index(es):
Relevant Pages
|
Loading
