Is there a rule of thumb for how many layers should be in a neural net? What about the weights initial range?

I have a dataset with one factor response and about 12 inputs, some factors. I've been using R's nnet to try and create a nn and have the two questions listed in the title. Thanks for any help.