Hi all,
I'm comparing the similarity between some samples and I'm not sure what's the best method to do it: correlation or distance.
A friend of mine has told me that distance has "more mathematic" in the background, but after some tests I don't see differences with correlation, any suggestion? Thanks in advance!
Dear Juan, in similarity analysis a correlation function should always be based on a distance or on a norm, therefore it is correct if you have not register differences.
Gianluca
Variance is a measure of dispersion. It measures the average distance of a variable from its mean. Covariance involves two variables. It means the average distance between two variables. Correlation is normalized covariance. The closer two variables are the more their correlation. For example the correlation between a variable with itself is maximum at +1.
But correlation is not exactly about the closeness but the association of two variables. Association is about proportional agreement of the variables on a point by point basis. You are therefore not completely right to think of correlation as a distance because close variables may not be associated closely.
First, you have to test whether there is a significant difference in the means. If not, you can use the appropriate measure of dispersion after excluding the outliers.
Thanks all for your help! I've found this paper that can shed light about this discussion... hope it clarifies!
Article Proximity Measures for Clustering Gene Expression Microarray...
- Badges
- Science topic
- Similar topics
- Mathematics