Re: how to compare the similarity of two groups?
- From: "Greg Heath" <heath@xxxxxxxxxxxxxxxx>
- Date: 15 Sep 2005 23:07:58 -0700
roger wrote:
> I have two groups: A, B, each group has several samples, such as
> a1,a2,...a8 in group A and b1, b2,...b8 in group B. and I have a set of
> genes, I have these genes expression profiles for every sample in A and
> B.
>
> could anyone please tell me how to compare the similarity (distance) of
> the two groups by the gene expression profiles of their samples?
>
> I am considering using euclidean distance as the metric, but I do not
> know what software and how can perform such analysis. The TIGR's GDM
> can give you a intuitive pictures, distances are represented by colors,
> however, if I can get a number of the average distance between two
> groups, that will be great!
> Thanks.
Euclidean, Mahalanobis, or Bhattacharya distance could be
appropriate if the groups are relatively separable via LDA
or QDA. Otherwise, I would measure similarity by the equal
error rate of a more general classifier,e.g., a neural network
(NN) or Support Vector Machine (SVM). Posterior probability
estimates from a NN allow the average log-likelihood ratio
to be used.
Hope this helps.
Greg
.
- References:
- how to compare the similarity of two groups?
- From: roger
- how to compare the similarity of two groups?
- Prev by Date: Re: How much skewness is too much?
- Next by Date: Re: Significance testing of Dependent proportions
- Previous by thread: how to compare the similarity of two groups?
- Next by thread: About rank invariant tests on ordinal data
- Index(es):
Relevant Pages
|
