Hi, I am new to research and statistics but have spent a lot of time reading textbooks and I understand the basics.

My question, is as follows. I have two predictor variables which are interval and discrete. They represent the number of friends (either followers or people that a user is following) Some respondents to my survey reported very high numbers and others low. These variables are skewed quite a bit. Now I understand that predictors don't need to be normal (it helps if they are) but my questions is when is it ok to transform variables. If I am correct the structure changes.. as my variables represent friendships you can't have 1.1 friendships. but in the same respect, I am using parametric tests.. such as linear regression so I am actually saying the variables are continuous. Do I leave them skewed or Log transform them because it will make the distribution much closer to normal.