Re: How to best deal with severely skewed data
- From: hrubin@xxxxxxxxxxxxxxxxxxxx (Herman Rubin)
- Date: 30 Apr 2008 14:27:41 -0400
In article <aeb411a0-2d6a-4805-914f-034485e6bc83@xxxxxxxxxxxxxxxxxxxxxxxxxxxx>,
padinora <padinora@xxxxxxx> wrote:
I would like to look at the interaction effect of 2 independent
categorical variables on continuous dependent scale scores (calculated
as means). The difficulty is that the data is not normally
distributed. The dependent variables are very skewed towards positive
responses (scale runs from 1-4, there are no 4 responses and very few
3s).
I tried several transformations (square root, log, negative
reciprocal, & rank) but still did not get a normal distribution. Does
anyone have suggestions on what I should try next? Any references
would be a huge help.
Thanks,
padinora
Normality of the DATA is NEVER important; normality of
the errors might be.
You have categorical data, not "scaled" data. Procedures
for analyzing categorical data should be used. There are
many of these which immediately come to mind; I am not
sure what your data looks like.
--
This address is for information only. I do not claim that these views
are those of the Statistics Department or of Purdue University.
Herman Rubin, Department of Statistics, Purdue University
hrubin@xxxxxxxxxxxxxxx Phone: (765)494-6054 FAX: (765)494-0558
.
- References:
- RE: How to best deal with severely skewed data
- From: padinora
- RE: How to best deal with severely skewed data
- Prev by Date: Within subject/cross subject design
- Previous by thread: RE: How to best deal with severely skewed data
- Next by thread: Operative characteristic curves and sampling
- Index(es):
Relevant Pages
|
