Re: discriminant analysis: How do I test for stat. significance?
- From: "Old Mac User" <chendrixstats@xxxxxxxxx>
- Date: 10 Nov 2006 06:11:18 -0800
You use SPSS "to do the statistical tricks." Really?? That's quite
interesting. If you are interested in doing magic, I suggest Dreamland
Magic at http://www.magical-tricks.com/index.html as a place to learn
more.
Now seriously, if you see the analysis of multivariate data as a form
of "doing tricks" then I suggest you get help from a professional
statistician. We see many posts here that begin with something like
"I'm using the SmartStats software to do.... (fill in the blanks)" when
it's obvious that the poster hasn't a clue. It's stuff like this that
gives "statistics" a bad name.
More and more people pump numbers through brand-name software untouched
by human brains.
One reply to your question (David Jones) suggested using a resampling
procedure. That was also the first thing that came to my mind.
Resampling is about as good as it gets when there's no nicely canned
"trick" at your fingertips. Virtually assumption free, it will blow the
minds of those who may perchance read your published paper.
Unfortunately, using resampling methods (which are really simple in
concept) requires some thinking and might turn out to involve work
outside of SPSS or any other available software. OMU
holg wrote:
Hmm, sounds interesting. Actually I use SPSS to do the statistical
tricks, since I´m not really good at statistics. Do you know of any
ready to use SPSS-functionality to perform the test? I want to make
sure that I quote the correct parameters because I want to publish the
results in a scientific paper.
Best wishes,
Holger
David Jones schrieb:
holg wrote:
Hi,SPSS
I have conducted a discriminant analysis using two separate sets of
independent variables for classification of 300 images falling into
three categories. How do I test for statistical significance when
tells me, that for the one set of parameters, 71% of images, for the
other 81% were classified correctly?
Best wishes,
Holg
If nothing else is available you could do the obvious. That is:
(i) choose a test statistic (possibly the % figure, possible some
measure of distance between categories);
(ii) evaluate a distribution for the test statistic under the null
hypothesis (which is presumably that categories can't be predicted
from the "parameters") ... do this by generating ramdomised sets of
data using your observed data as the basis of this ... randomly
reassign images to categories, keeping the numbers in each category
the same ... for each randomised dataset reapply the discriminant
analysis and extract the test statistic for each. You may only need to
do 10 or so randomised datasets to get a good enough idea of the
significance of your original test statistic, or more if you need to
do some formal reporting (100-200 say).
David Jones
.
- References:
- Prev by Date: Re: discriminant analysis: How do I test for stat. significance?
- Next by Date: Russian education system observe
- Previous by thread: Re: discriminant analysis: How do I test for stat. significance?
- Next by thread: Russian education system observe
- Index(es):
Relevant Pages
|
|
