On Mon, 3 May 2004, MrIlfar wrote in part:
> Can I simply calculate the joint probability as:
>
> P = (400*0.9936*1.0 + 596*0.9936*0.9933 + 395*0.9936*0.86 ) /
> (400+596+395)
> P = 95.12 ???
I wouldn't think so. OTOH, I don't know what you think you want to mean
by "the joint probability of a hierarchical classifier system". And
while I _think_ I know which way round you've represented the "confusion
matrix" for C1, you haven't supplied explicit information.
So if you supply some adequate definitions, I'll be happy to comment on
your reasoning (although I think the lack of confidence you report is
with respect to probability, not statistics). -- DFB.
On Mon, 3 May 2004, MrIlfar wrote:
> Hi, hope someone can comment on my reasoning (due to lack of
> confidence in statistics) on this probability problem:
> I want to find the joint probability of a hierarchical classifier
> system. It consists of 4 individual classifiers C1, C2, C3 and C4:
>
> C1
> |
> ---------
> | | |
> C2 C3 C4
>
> C1 classifies between groups {a,b},{c,d,e} and {f,g}
> C2 classifies between a and b
> C3 classifies between c, d and e
> C4 classifies between f and g
>
> So, in order to classify an item (e.g. f) correctly it is required
> that C1 classifies correcly such that the next classifier to be
> invoked is C4 (that classifies between f and g).
>
> I have the individual classification rates and the confusion matrices
> for each classifier - but I am not sure how to compute the "joint"
> probability..
>
> The individual classification rates are as follows:
>
> C1: classification rate: 99.36%
> Confusion matrix:
> 400 0 0
> 0 596 4
> 0 5 395
> (as you see there are 400 of type {a,b}, 600 of {c,d,e} and 400 of
> {f,g}.)
>
> C2: classification rate: 100%
> C3: classification rate: 99.33%
> C4: classification rate: 86.0%
>
> Can I simply calculate the joint probability as:
>
> P = (400*0.9936*1.0 + 596*0.9936*0.9933 + 395*0.9936*0.86 ) /
> (400+596+395)
> P = 95.12 ???
> I would appreciate any comments on the correctness of this :)
>
> Many Thanks
>
> McIlfar
------------------------------------------------------------
Donald F. Burrill [EMAIL PROTECTED]
56 Sebbins Pond Drive, Bedford, NH 03110 (603) 626-0816
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
. http://jse.stat.ncsu.edu/ .
=================================================================
