数理统计Etimation.pptVIP

* * Population mean ? (height, weight …) Sample mean (Point) estimate Estimation * Interval Estimates From a sample of 200 cartons we find a sample mean weight of 375g and a sample SD of 30g. How confident are we that this estimate (sample mean) is within 2g of the true value (population mean)? A: It would matter if the sample was small. But for large samples the sample SD (s) gives a pretty good estimate of the population SD (σ), so it makes little difference whether we have s or σ. Q: We have only a sample SD, and we don’t know the population SD. Is this a problem? Example: * = p( -0.94 ? Z ? 0.94 ) = 0.6528 We can be about 65% confident that our sample mean is within 2g of the population mean. * We can be about 65% confident that the population mean is between 373g and 377g. = 0.6528 We can be about 65% confident that our sample mean is within 2g of the population mean. We worked this out without using any particular value for the population mean. Note: We can turn the statement around: if: is within 2g of μ, then: μ is within 2g of . * Confidence Interval for Mean (large samples): The sampling distribution lets us work out our interval in SEs either side of the mean. 0.95 -1.96 1.96 Confidence intervals start with a confidence level. This is often 95%, but 90% and 99% are also used. When estimating a population mean ? using from a large random sample, we can be 95% sure that will be within 1.96 SEs of ?, and hence that ? will be within 1.96 SEs of . * More compactly, we say that the limits of a 95% confidence interval for ? are given by – A lower boundary of and an upper boundary That is, we can be 95% sure that ? lies in an interval defined by – 0.95 -1.96 1.96 * More compactly, we say that the limits of a 95% confidence interval for ? are given by – A lower boundary of and an upper boundary That is, we can be 95% sure that ? lies in an interval defined by – 0.95 -1.96 1.96 * Again, since 99% of values