1. I. Survey Design Basics

2. A. Foundations What is your idea or argument? – Ex. Public anger about the ACA will hurt the Democrats in 2014. What does that argument imply about data (your hypothesis)? – Democrats will do worse than expected or normal – Their underperformance will be accounted for by public attitudes toward ACA.

3. B. Conceptualization What concepts are you trying to measure? – What? Toward whom? When? – Electoral support (what) for Democratic candidates (who) in the November 2014 election (when) What Quantities are you trying to measure? – Averages: Means, Percentages (expectations, sizes of groups in society) – Quantiles: Medians, quintiles, etc. (value of X such that Q percent are less than X). (inequalities) – Variance: Spread (risk) – Relationships: correlations, differences in means, regressions (association, causation, prediction)

4. C. Population Definition. Universe of all persons (or units) you seek to study. Finite and infinite. Finite: known, fixed population. All people in US today. Infinite: Continuous variables, Future (distribution). Stock market value tomorrow.

5. D. Sample Construction Mode of Contact. How communicate. – In person, mail, phone, internet Who is contacted? – Random – Representative

6. E. Survey Instrument Means of collecting information Question Format Constraints – time limits, change behavior by asking too much.

7. F. Examples 1.Exit Poll Questionnaire: 18 questions Sample Precincts (problem of clustering) Sample individuals as leave Respondents and Non-Respondents Device (paper, handheld?) 2.Phone Polls Questionnaire about 20 or so questions Random Digit Dialing Very high non-response (what’s random?)

8. II. Thinking About Data

9. Questions, Frequencies and Tables DemocratRepublicanIndependentOthers 35.3% (4,866) 24.4% (3,360) 26.7% (3,687) 13.7% (1,888) In politics today, do you consider yourself to be a Democrat, Republican, Independent, or something else?

10. Race and Hispanicity White Alone (H and NH) Black Alone (H and NH) Asian Alone (H and NH) HispanicWhite Non-Hisp White Hisp 77.9%13.1%5.1%16.9%63.0%14.9% WhiteNon-White Hispanic14.9%2.0% Non-Hispanic63.0%20.1%

11. Census Version: Race Question Black (13%) White (79%) Asian (5%)

12. Census Version: Hispanicity Question NONHipsanic (83%) - Hispanic (17%)

13. Census Version: Race and Hispanicity are Separate Question Black (13%) Hipsanic (17%) White (79%) Asian (5%)

14. Tabular Presentation Race and Hispanicity Non-WhiteWhiteTotal Hispanic2.0% 11.8% 9.0% 14.9% 88.2% 19.0% 16.9% 100.0% - Non-Hispanic20.1% 22.6% 91.0% 63.0% 77.4% 80.9% 89.1% 100.0% - Total22.1% - 100.0% 77.9% - 100.0%

15. Marginal and Joint Frequency Terminology – Variables Y = i, i = 1, 2, … I X = j, j = 1, 2, … J – Marginal Frequencies or Probabilities. P(Y=i) or P(X=j) – Joint Frequencies or Probabilities. P(Y=i and X = j)

16. Tabular Presentation Race and Hispanicity Non-WhiteWhiteTotal Hispanic2.0% 11.8% 9.0% 14.9% 88.2% 19.0% 16.9% 100.0% - Non-Hispanic20.1% 22.6% 91.0% 63.0% 77.4% 80.9% 89.1% 100.0% - Total22.1% - 100.0% 77.9% - 100.0%

17. Tabular Presentation Marginals on Hispanicity Non-WhiteWhiteTotal Hispanic2.0% 11.8% 9.0% 14.9% 88.2% 19.0% 16.9% 100.0% - Non-Hispanic20.1% 22.6% 91.0% 63.0% 77.4% 80.9% 83.1% 100.0% - Total22.1% - 100.0% 77.9% - 100.0%

18. Tabular Presentation Race and Hispanicity Non-WhiteWhiteTotal Hispanic2.0% 11.8% 9.0% 14.9% 88.2% 19.0% 16.9% 100.0% - Non-Hispanic20.1% 22.6% 91.0% 63.0% 77.4% 80.9% 83.1% 100.0% - Total22.1% - 100.0% 77.9% - 100.0%

19. Tabular Presentation Joint Frequencies for Race and Hispanic Non-WhiteWhiteTotal Hispanic2.0% 11.8% 9.0% 14.9% 88.2% 19.0% 16.9% 100.0% - Non-Hispanic20.1% 22.6% 91.0% 63.0% 77.4% 80.9% 83.1% 100.0% - Total22.1% - 100.0% 77.9% - 100.0%

20. Thinking Conditionally Definition – A Conditional Statement is the Set of Values of a variable (say Y) subject to the restriction that another variable or variables take on a specified set of values. Terminology. – Y given X=j or Y|X=j – Example: Y=Hispanic|X=Non-White. – That’s different from Y= Hispanic and X = Non-White. How So? How is a statement Y given X different from Y and X?

21. Conditional Frequency or Probability, Defined P(Y=i|X=j) = P(Y=i and X = j)/P(X=j) P(X=j|Y=i) = P(Y=i and X = j)/P(Y=i) In statistical software these are called row and column percentages in tables. Joint frequencies are called cell frequencies tab y x, row col cel

22. Census Version: Race and Hispanicity are Separate Question Black (13%) Hipsanic (17%) White (79%) Asian (5%)

23. Census Version: Race and Hispanicity are Separate Question Black (13%) Hipsanic (17%) White (79%) Asian (5%) W and H

24. Ex. Race and Hispanic Marginal Probabilities Joint Probabilities – Hispanic = Yes:.17H, W:.15 – Hispanic = No:.83H, NW:.02 – White = Yes:.78NH, W:.63 – White = No:.22NH, NW:.20 Conditional – P(W|H) =.15/.17 =.88 – P(H|W) =.15/.78 =.19

25. Example 2. Race and Party (CCES 09) DemocratRepublicanIndependentOtherTotal White3,037 29.8 62.4 2,932 28.8 87.3 2,897 28.5 78.6 10,179 100 73.8 Black1,143 70.8 23.5 65 4.0 1.9 241 14.9 6.5 1,615 100 11.7 Hispanic472 40.6 9.7 219 18.8 6.5 240 20.6 6.5 1,163 100 8.4 Other843 100 6.1 Total4,866 35.3 3,360 24.4 3,687 26.7 1,887 7.8 13,800