The 10, 5, and 1 percent levels of significance are the most common for testing the statistical significance of individual regression coefficients. However, any level of significance chosen for a confidence interval or test of significance exposes you to type I and type II errors.