Center Home -> Content Areas Home -> Math Home -> Project Activities -> Calculator Activities ->

Survey Sample Size and Confidence Intervals

Activity Description Activity Guide


On Sunday, March 12, the Charlottesville Daily Progress contained an article reporting the results of a Washington Post-ABC News Poll. The article reported that "Gore leads Bush by 48 percent to 45 percent, a statistical tie. It also reported "A total of 1,218 adults, including 999 self-described registered voters were interviewed" and the "margin of error for the overall results is plus or minus 3 percentage points, and slightly larger for results based on only the registered voters."

Part 1: Determining and Interpreting a "Margin of Error"

(1) Interpret, in your own words, what is meant by a "margin of error of plus or minus percentage 3 percentage points " and what is meant by " a statistical tie."

(2) What kind of statistical test is appropriate for determining the confidence interval for such a survey? Explain.

(3) Use you graphing calculator to assess the accuracy of the newspaper article's statement about the margins of error for these samples. Was the newspaper reporting accurate? What unreported confidence level was used to determine the margin of error? [95%]

(4) Again interpret "margin of error," this time incorporating the confidence level.

(5) Predict what the "margin of error" would be for this proportion of voters at the 99% confidence level. Calculate the "margin of error" for the 99% confidence levels, and discuss your prediction. In general, discuss how a survey's margin of error is related to the confidence level chosen.

Part 2: The Effect of Sample Size on Confidence Intervals

(6) The survey reported in the newspaper was based on samples of 1218 and 999 respondents. Discuss the appropriateness of this sample size for obtaining reliable results in a presidential election.

(7) Predict what the 95% confidence level "margin of errors" would be for surveys, with the same proportion, but with sample sizes of n=100, 200, 500, 1000 and 10,000.

(8) Use your calculator to find the margins of error for these samples.

(9) Discuss your predictions, and the gains and trade-offs involved with various sample sizes. Discuss reasonable sample sizes for presidential elections.

(10) Write the general expression for the standard error for this distribution. Examine this expression and predict the shape of the graph of this standard error as a function of sample size. Discuss how this predicted shape is consistent with the results found for Task 8.

Note: The population proportion p is estimated by the sample proportion x/n, with standard deviation s = Ö (p(1-p)/n).

(11) Using the proportion for Gore and the 95% confidence level, graph the standard error as a function of sample size, and trace along the graph. Compare your observations with the results found in Task 8.

(12) Predict the position of the graph of the 99% standard error as a function of sample size.

(13) On the same screen, graph the 99% confidence level standard error as a function of sample size. Discuss your observations.

Part 3: The Effect of Proportion on Confidence Intervals

In the above tasks sample sizes were analyzed and graphs were drawn for margins of error at the 95% confidence level when the sample proportion was known to be .48. How do the margins differ if the sample proportion is something other than .48? In other words, can one use the analysis from above to determine an appropriate sample size when the proportion is not already known?

(14) Consider the standard error expression as a function of proportion for a fixed sample size. Describe how the margin of error depends on the proportion? Predict the graph of this function.

(15) Graph the 95% standard error as a function of the proportion for a sample size of 1000. Analyze and trace this graph, and discuss the effect of proportion on margins of error and how this would affect a pollster's determination of an appropriate sample size.

Extension 1: The 2000 Presidential Primaries

An earlier Dartmouth College-Associated Press poll conducted by during the 2000 presidential primary season yielded the following data. For the Republicans, Senator John McCain received 42% and Governor George W Bush received 33% of the votes from a sample of 518 likely voters. For the Democrats, Vice President Gore received 48% and former Senator Bradley received 41% of the votes cast by 418 likely voters. The newspaper reported margins of error of 4.5 and 5 percentage points, respectively. Comment on these numbers.




Back to Project Activities | Back to Math Homepage

Send questions or comments here.
Last modified on August 14, 2001.