Introducing Undergraduates to Concepts of Survey Data Analysis

Figures & data

Table 1 Summary table showing the results from the hypothetical university survey.

	Yes	No	Total
Undergraduate	30	70	100
Graduate	80	20	100
Total	110	90	200

Fig. 1 Screenshot from the Should It Pass? ShinyApp that allows students to modify the hypothetical survey design by adjusting the sample sizes of each strata. Students can also modify the proportion in each group that support the resolution. Graphs display the unweighted sample data and the weighted population estimates allowing students to compare the unweighted results and weighted estimates for the specified situation. This app as well as corresponding datasets and student lab handout are freely available at https://stat2labs.sites.grinnell.edu/weights.html.

Table 2 First six rows of the university survey data demonstrating the use of weights.

ID	Status	Vote	Weight	Weighted vote
1	Undergraduate	0	200	0
2	Graduate	1	10	10
3	Undergraduate	1	200	200
4	Undergraduate	1	200	200
5	Graduate	0	10	0
6	Graduate	1	10	10
$\dots$	$\dots$	$\dots$	$\dots$	$\dots$
Sum		110	21,000	6800

NOTE: The sum of each column based upon all 200 cases in the sample is also shown.

Fig. 2 Screenshot from the interactive Political Preferences 1 ShinyApp which demonstrates that results weighted by race have much higher estimated Republican responses than the unweighted sample data. The variable displayed on the x-axis and the subgrouping variable can be changed to explore the dataset. The results can be shown in counts, as selected in the screenshot, or in percentages. This app as well as corresponding datasets and student lab handout are freely available at https://stat2labs.sites.grinnell.edu/weights.html.

Table 3 The percentage of students who were able to properly complete the activities in the pre- and post-quiz related to weighted data.

Questions	Introductory statistics		Statistical modeling
	Pre ($n = 42$ )	Post ($n = 17$ )	Pre ($n = 32$ )	Post ($n = 22$ )
(1) Calculate unweighted estimates.	93	88	97	100
(2) Calculate weighted estimates.	24	50	62	84
(3a) Identified that unweighted estimates would lead to incorrect results.	61	75	91	95
(3b) Identified that the sample did not properly represent the population.	43	59	73	93

CBS News and The New York Times (2012), CBS News/New York Times Monthly Poll #4, October 2010, Ann Arbor, MI: Inter-University Consortium for Political and Social Research [distributor], DOI: 10.3886/ICPSR33183.v1.

 Google Scholar