*****  To join INSNA, visit http://www.insna.org  *****

If the values are all EQUAL to zero (as opposed to many different values near to zero), then no transformation in the world will help, because all the zero values will be transformed to the same new value. For example, with a log transformation, say you add a constant to the zero values (because you can't take the log of zero). Let's say you add 10 (to make the arithmetic easy), and take a base 10 log. Then all of those zeroes are now ones. You still have a spike, but now at one instead of zero. If you've got an overly long tail, a log transformation will bring them in (helping with skew) but if you're "skewed" because of a spike at zero, a transformation will not "unspike" them. 

Apologies if I did not correctly understand your issue. 

Sent from my iPhone - Frank

> On Jun 9, 2017, at 9:57 AM, Chao yawo <[log in to unmask]> wrote:
> 
> ***** To join INSNA, visit http://www.insna.org *****
> Hello,
> 
> Good morning.  I am writing to ask for some advice on possible transformation strategies for a 
> network density variable that serves a a predictor (IV) in a logistic regression model.
> 
> The distribution is extremely skewed (with a lot of zeros) - in fact the mean approaches 0.  but I believe the zeros are meaningful, reflecting extremely sparse network, and their effect need to be captured in the model. 
> 
> What are some of the ways networkers deal with skewed distributions.,  I can categorize the variable (I am hesitant to do that, for well known cautions against categorizing interval level measures).  A second option is to transform it (e.g., log it).   But I am wondering if there is any other way to deal with this issue or the advisability of the two options I am considering.
> 
> thanks in advance,
> 
> Cy
> _____________________________________________________________________ SOCNET is a service of INSNA, the professional association for social network researchers (http://www.insna.org). To unsubscribe, send an email message to [log in to unmask] containing the line UNSUBSCRIBE SOCNET in the body of the message.

_____________________________________________________________________
SOCNET is a service of INSNA, the professional association for social
network researchers (http://www.insna.org). To unsubscribe, send
an email message to [log in to unmask] containing the line
UNSUBSCRIBE SOCNET in the body of the message.