statistical difference between normal and non-normal data

Here's a table showing you why you might choose each test. It sounds like you should use a non-parametric test if you know some of the data is non-normal.

With a sample size of 100 you should be able to use a parametric test so long as you have reason to assume that the population data "should" be normal. It depends on what your variables are and how they are naturally distributed. Take blood pressure readings, for instance. There's no reason to assume that bp readings aren't normally distributed, even if your test data looks skewed. OTOH, it's never wrong to use a parametric test. You'll lose some power but you don't run the risk of violating your assumptions. Some parametric analyses are more robust than others when it comes to lack of normality. It all depends on your hypothesis and the distribution of the underlying population, especially with a large sample.

er, that should read that it's never wrong to use a non-parametric test.