Math 37 - Lecture 16
Confidence Intervals (6.1, 8.1)
Goal: Make a conclusion about the population and use probability to express the strength of our conclusion.
Confidence Interval - Estimate the value of the population parameter.
Procedure:
If we took many samples from the population, how would the statistic vary?
E.g. If the statistic follows a Normal distribution, can say that 95% of the time the statistic will be within two standard deviations of the parameter. Thus, in 95% of samples, the parameter will be within 2 standard deviations of the statistic.
Proportions - The sampling distribution of
is approximately Normal (Normal approximation to the Binomial) with mean p and standard deviation
. In 95% of samples, p will be in the interval (
-2
,
+2
).
Estimate SD(
) by plugging in
for p.
Example In June 1994, the cable television show Nick-at-Nite conducted a phone-in survey of its viewers asking: "Who had more magic powers - Jeannie or Samantha"? Samantha received 810,000 votes to Jeannie's 614,000.
- What's the population parameter?
- Find a 95% confidence interval for the population parameter
- What do you think about the width of this interval? Does this make sense?
Confidence interval indicates plausible values for the parameter based on the observed data.
Means - The sampling distribution of
is Normal (if population is) or approximately Normal (Central Limit Theorem) with mean m an standard deviation
. In 95% of samples,
will be within 2
of m. In 95% of the samples m will be within 2
of
, that is, in the interval: .
Example In a 1993 study, researchers took a sample of 25 people who claimed to have an intense experience with an unidentified flying object (UFO) and a sample of 25 people who did not claim to have such an experience. They then compared the two groups on a variety of variables, including IQ. The sample mean IQ of the UFO group was 101.6 and the standard deviation of these IQ's was 8.9
- Construct a 95% confidence interval for the population mean IQ of all people who have had intense UFO experiences.
- Explain why the confidence interval is not so narrow.
What if we wanted to be 90% confident? 97% confident?
In general - For any Normal distribution, given any value C, can find a number, z*, such that the Normal distribution has probability C lying within z* standard deviations of its mean: use Table A to find z* such that the interval (-z*,z*) has probability (1-C)/2 on either side.
Critical Value, z*
Let z* be the point in Standard Normal distribution with probability p lying to its right. Call z* the (upper) critical value.