[R] How to deal with zero-inflated proportion data?

Stefan Sobernig stefan.sobernig at wu.ac.at
Fri Feb 15 15:09:03 CET 2013

Dear List,

In my previous posting 
(https://stat.ethz.ch/pipermail/r-help/2013-February/347593.html), I 
refer to playing around with distribution fitting for a particular sort 
of data (see dat1 in the previous posting):

I collected absolute node degrees of some network structure and computed 
the relative node degrees. This is because I want to compare different 
networks at another stage.

Based on relative degrees, I attempted some distribution fitting 
procedures (see my previous posting). However, I realized that zero 
values (representing isolates in the network) in an otherwise continuous 
variable are problematic (well, not much of a surprise).

I am wondering how to best deal with those isolates/zeros? Excluding 
them entirely would be an option but isolates are relevant for my 
analysis. Are there best practises for dealing with such a hybrid variable?

I am aware of censored methods (for regression analysis etc.), but not 
when it comes down to distribution fitting, I fear.

Many thanks,

More information about the R-help mailing list