[Gator67]

This is cool Dave. To get a feel for the bounds of error given your sample size, take the square root of the following quotient: (percentage of a certain color*1-percentage of a certain color)/n. This gives you a standard error of the proportion. Then multiply by the value from the normal distribution at the confidence level your interested in (z). Then subtract (and add) this z value to the percentage of the color you're interested in. This will give you the lower and upper bounds of the percentage at the confidence level you chose. So for Linden green (.041*.959)/317=.00012403. The square root is .01113709. If I want 95% confidence I multiply by 1.96, which gives me .0218287. I subtract and add that number to .041 and I get 1.92 to 6.28. This means I'm 95% confident that the percentage of Linden Green GTOs falls between 1.92 and 6.28, or between 1,567 and 5,134 cars. You'll see that if you want to be less confident, the interval will narrow. You'll also see that the standard error will be wider the higher the percentage--probably closer to 4% for the top three colors. If you want to make comparisons between colors, look to see if the confidence intervals overlap. I suspect you'll find three bands at the 95% confidence level.