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 have a list of tokens/words which co-occur with each other. Some of these tokens co-occur more than the others. Moreover, there's a conditional relationship between many. For example, "bat" occurs conditionally with "sport" & "baseball". Such conditional relationships can be on multiple variables. I want to represent these "conditional relationships" (depending on presence/absence of values) using a DAG. Is a graphical model the correct way to do so?

Mathematically, I have a large symmetric matrix r x c. The values of this matrix are binary signifying the collocation of r,c viz. if r,c = 1 it means that r,c are co-occurring and may have a dependence between them.

share|improve this question
    
What kinds of analyses or decisions do you want to make about these relationships? –  whuber Jan 9 '12 at 21:26
    
The rows/columns are in principle words while the values of the matrix represent co-occurrence of words. I want to infer/reason the relationship between the words preferably as a directed acyclic graph. –  Dexter Jan 9 '12 at 21:34
    
The matrix is mathematically equivalent to the DAG. This returns us to the crux of the matter: what do you want to do with it? Merely draw a picture? Document the data? If you want to go beyond such things, it would help for you to be more specific about what you mean by "infer" or "reason" a "relationship." –  whuber Jan 9 '12 at 22:41
    
Pardon me for making this complex than it already is.Lets go back to square one. I have a list of tokens/words which co-occur with each other. Some of these tokens co-occur more than the others. Moreover, there's a conditional relationship between many. For example, bat occurs conditionally with sport & baseball. Such conditional relationships can be on multiple variables. I want to represent these "conditional relationships" using a DAG. Is a graphical model the correct way to do so? –  Dexter Jan 10 '12 at 7:31
1  
It is always helpful to go back to a statement of the problem that really concerns you. Why don't you edit the question to reflect this last comment? You might also consider indicating what you intend to do with this representation of the word associations, because that can point the way to good answers. –  whuber Jan 10 '12 at 13:51

1 Answer 1

You may be looking for the log-likelihood ratio: http://tdunning.blogspot.com/2008/03/surprise-and-coincidence.html

share|improve this answer

We're looking for long answers that provide some explanation and context. Don't just give a one-line answer; explain why your answer is right, ideally with citations. Answers that don't include explanations may be removed.

Your Answer

 
discard

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.