For large random samples, when we look at sample means or proportions, what happens to the width of a confidence interval when the confidence level is increased from 68% to 95% (holding all else constant)? Explain briefly.