Sorry this took me so long to write up a summary, but I didn't forget:)
I was looking to figure out how to do a test with small very small numbers.
What I came up with is that if I had whole numbers for my observed and
expected values, I would have been able to do a Fisher exact test in SAS
very easily, or in StatExact. The number 5 as the smallest expected is a
very conservative number and so since most of my values are above that and
only one or two are below, and most of those are above 3, it is OK to do the
Fisher.
However, since I do not have whole numbers, at least in SAS, I can't do the
Fisher. However, a chi-square is a reasonable estimate and will work fine.
I found an on-line place to do a Fisher without using whole numbers and my
values were very close to those I found using the chi-square in SAS. I also
tried to use the Stat-Exact sample but my overall sample size is too large
and I couldn't make it work and I haven't had a chance to "buy" their
short-term demo.
Thank you to everyone who was very helpful in giving me some suggestions,
Oh, and to the person who was worried about my cutting a continuous variable
into pieces, I was actually using different types of area (i.e. soil types)
and so that was not an issue. But thanks for the concern.
Ingrid
Ingrid Friberg, MHS
Epidemiologist
Infant Botulism Treatment and Prevention Program
California Department of Health Services
2151 Berkeley Way, Room 506
Berkeley, CA 94704
Phone: 510-540-2646; Fax: 510-540-3205
email: [log in to unmask]
|