The Math Forum

Search All of the Math Forum:

Views expressed in these public forums are not endorsed by NCTM or The Math Forum.

Math Forum » Discussions » sci.math.* »

Notice: We are no longer accepting new posts, but the forums will continue to be readable.

Topic: discriminant analysis: How do I test for stat. significance?
Replies: 1   Last Post: Nov 10, 2006 5:15 AM

Advanced Search

Back to Topic List Back to Topic List Jump to Tree View Jump to Tree View  
David Jones

Posts: 637
Registered: 12/7/04
Re: discriminant analysis: How do I test for stat. significance?
Posted: Nov 10, 2006 5:15 AM
  Click to see the message monospaced in plain text Plain Text   Click to reply to this topic Reply

holg wrote:
> Hi,
> 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

Point your RSS reader here for a feed of the latest messages in this topic.

[Privacy Policy] [Terms of Use]

© The Math Forum at NCTM 1994-2018. All Rights Reserved.