According to the latest census, there are 200 million Americans aged 20 and over. You want to estimate the number of Americans in this age group who are without health insurance. You plan to select a random sample and use sample results to build a 90% confidence interval. You want a sample size that will give an interval no wider + 1 million people. If a preliminary estimate of the number of uninsured Americans aged 20 and over is 40 million, how large a sample would you recommend?