Take the 2-minute tour ×
Cross Validated is a question and answer site for people interested in statistics, machine learning, data analysis, data mining, and data visualization. It's 100% free, no registration required.

I am working on Gene-Gene interaction networks. I build a graph by adding edges between genes (nodes) representing statistical interaction in prediction of a quantitative parameter value (say, brain volume) in a multiple regression model.

For creating a well-connected graph, I have lowered the p-value threshold of the linear model interaction term in order to include as many edges as possible, so it is likely that a proportion of these links are false-positives (actually I have omitted multiple comparison correction and used a nominal p<0.05).

My question is that, if some nodes show a very high degree centrality‎ which is very unlikely to happen in a random Erdős–Rényi model (aka Poisson distribution), can I make a statistical inference that this node is biologically relevant to the brain size parameter because it has "received" more than random edges, even if the whole graph does not show a Erdős–Rényi degree distribution pattern? What is the best test for statistical evaluation of this hypothesis, and if there exists one, is it also necessary to perform a multiple-comparison?

share|improve this question
    
Do you mean to say that the edges represent interactions which are statistically significant as determined by their use in a regression model? –  learner Mar 5 '13 at 21:04
    
@learner Yes the edges are GLM interaction terms, but are not significant at a multiple comparison threshold, i.e. we have N nodes and compute Nx(N-1)/2 interaction terms. –  Sourena Mar 5 '13 at 21:36
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Browse other questions tagged or ask your own question.