 To test to see if you can use the normal approximation to the binomial distribution, you need to check two things. If N, which is the number of the sample, times the proportion, P is greater than 5, and also that Q times N is greater than 5. Q, of course, is just 1 minus P. I've set up this little table in Excel. Here's an N of 1,000, and we're given a P of 0.036, which is 3.6%. Q is 1 minus P, 0.964. And N times P is 36. That's greater than 5. Here I've used some conditional formatting to color those cells green if it's okay and red if it's not. If my sample size is only 100, you can see that N times P is now less than 5, which means it's not okay to use the normal approximations of binomial, and you must use the binomial to determine probabilities.