That is not an easy question to answer, since it is very dependent on what I am doing and how I am going about my experiments. I am a nuclear physicist, so I wind up using Monte Carlo codes a lot, and the accuracy of a Monte Carlo code is very much dependent on using actual random numbers, which is not really possible with computers, and takes too much time to do using tables and such like they used to do back in the 40's, but we can get very close. Using proper data fitting techniques (like the least squares method and such) is also important. I can't really get too specific since the statistical methods I use can and do change and depend on what I am doing. I don't have any favorite or preferred methods, I use what I have to.