Foundations of Research 13. Survey Research Part 1 This is a PowerPoint Show Open it as a show by going to “slide show”. Click through it by pressing any key. Focus & think about each point; do not just passively click. © Dr. David J. McKirnan, 2014 The University of Illinois Chicago McKirnanUIC@gmail.com Do not use or reproduce without permission Center for Epidemiologic Studies National Institute of Mental Health 1 Foundations of Research 13. Survey Research Part 1 We will address five topics in the Survey modules: Part 1 will address the first four. Topic areas & formats General issues in Survey research Sources of bias (or fraud…) in survey research Examples of surveys Testing Hypotheses with surveys 2 Foundations of Research Survey Research ✓ Topic areas & formats General issues in Survey research Sources of bias (or fraud…) in survey research Examples of surveys 3 Foundations of Research What do surveys measure? 4 Knowledge Information re: current events, political or consumer choices Awareness: e.g., of Public health resources, government decisions... Attitudes and Beliefs Preferences or evaluations: e.g., attitudes toward racial or ethnic groups, consumer preferences... Beliefs about political or social events: “which party provides the strongest security for the U.S….?” Feelings or moods: quality of life, depression / anxiety, marital satisfaction... Behavior Behavioral intentions; Intent to vote, financial plans, exercise goals. Self-reports of previous or on-going behavior; voting in the last election, alcohol and drug use, exercise patterns. Foundations of Research Survey research; General uses of surveys Survey methods have a wide range of applications, from single-item consumer satisfaction (“How useful did you find this web site”) to full-fledged, theory-driven behavioral research. For convenience we will consider 5 categories: Descriptive research Testing hypotheses; Testing the generalizability of experimental results; Predicting an event or outcome; Pragmatic / applied questions. 5 Foundations of Research Uses of surveys; descriptive research 6 Descriptive research Epidemiology is the study of how behaviors, disease states, or similar issues are distributed across the population. Epidemiology uses many methods, such as standard crime or disease reporting. Even Google search data are used to track flu spread. Many epidemiological studies use direct survey methods, such as phone or face-to-face survey interviews. Knowledge of, e.g., how to access health care… Feelings or moods, such as the rate and distribution of depression… Behavioral patterns, such as alcohol or drug use or gun ownership… Foundations of Research The origins of epidemiology 7 In London of 1854 a cholera outbreak raged through several poor neighborhoods of London. Sewage and other effluvia that ran through gutters created a dreadful smell (a “miasma”) that was blamed for the outbreak. The concept of infectious disease transmission through water supplies was not well understood. Dr. John Snow, one of the physicians charged with stopping the epidemic, noted a particularly fetid cesspool in front of 40 Broad St., proximal to a water pump used by the neighborhood. He decided to empirically map the cholera cases in the area. He proposed that water from the pump, not the miasma, was the cause of the outbreak. He was was generally disbelieved, but convinced the town governors by his evidence. Image: http://leabridge.org.uk/waterworks/timeline/1 852-1880.html Only much later would tracking of disease outbreak be labeled ‘epidemiology’. Click for Wikipedia article. Dr. John Snow’s Cholera map & the closing of the Broad St. pump. Foundations of Research 8 The origins of epidemiology Dr. John Snow’s Cholera map & the closing of the Broad St. pump. of cases to be near the pump at 40 Broad St…. Click for Ted talk by Steven Johnson. …and to radiate out from there. As he noted in his 1855 book: "I had an interview with the Board of Guardians of St. James's parish, on the evening of Thursday, 7th September, and represented the above circumstances to them. In consequence of what I said, the handle of the pump was removed on the following day.” By carefully describing the distribution of cases and the circumstances around the pump, Snow was able to empirically demonstrate a likely cause. His hypothesis was supported by the epidemic quickly subsiding once the pump handle was removed. S. Johnson, The Ghost Map (2007), Riverhead Books. Snow’s map showed the bulk Foundations of Research Uses of surveys; descriptive research 9 Descriptive research Epidemiology Political / social description is what we often think of as surveys. Opinion polls about society, the government, or current events, e.g. Gallup Polls, or systemic studies by Pew Memorial Trust. The Consumer Confidence Index is a highly standardized poll that is used for basic economic decision making. The Census, of course, is our national information source. Foundations of Research Uses of surveys; descriptive research 10 Descriptive research Epidemiology Political / social description. Testing hypothesis Assessing blocking variables We often assess blocking variables to test how a given attitude or behavior varies across important social groups. e.g., gender, age group, ethnicity, geographic location… Correlational studies A key form of analysis is examining the association among different variables e.g., what are the correlates of dieting… Foundations of Research Uses of surveys; descriptive research Descriptive research Epidemiology Political / social description. Testing hypothesis Assessing blocking variables Correlational studies Examine generalizability of experimental results E.g., The Consumer Reports survey on therapy we discussed in quasiexperiments Predict event or outcome; E.g., election polling Pragmatic / applied, E.g., marketing, or consumer surveys. 11 Foundations of Research Surveys; populations Who do we want to generalize to? Our sampling frame is based on our hypothesis or empirical question. Sampling: breadth internal validity tradeoff Key dimensions: Demographic ethnic / age / gender groups, “all Americans”… Behavioral “likely voters”, alcohol users, home buyers... Self-identification Republicans / Democrats, “students”… See Design and sampling overview See also: diminishing validity of political polling. 12 Question Formats: Closed-ended Foundations of Research items Specific rating scale or highly structured prompts Most reliable for concrete behaviors An attitude can be assessed in several ways: Direct (face valid) assessment Research methods is a wonderful course… Does not 1 2 3 4 5 6 7 strongly agree at all agree Behavioral (content valid) indictors How many times this semester have you skipped class? How many hours per day do you spend reading the material? 13 0 ______ 2.5 ______ Researchers typically use the M of several related items to create a more reliable measure of a variable. Foundations of Research 14 “Closed-ended” items, cont. Example: The CES-D Depression inventory: We may use the M score of these 9 items as our depression index… Moods & Feelings Below is a list of different feelings. Circle the number that shows how many days you felt each of these over the PAST WEEK. I was bothered by things that usually do not bother me. I felt I could not shake off the blues even with help from my friends or family. I had trouble keeping my mind on what I was doing. I felt depressed. I felt that everything I did was an effort. My sleep was restless. I was happy. I enjoyed life. I felt sad. Rarely or none of the time A Little of the Time A moderate amount of the time Most or all of the time (less than 1 day) (1 or 2 days) (3 - 4 days) (5 - 7 days) 0 1 2 3 0 1 2 3 2 3 2 2 2 2 2 2 3 3 3 3 3 3 0 1 Sum of item ratings /19 0 0 0 0 0 0 1 1 1 1 1 These items are “reversed” in the final score Foundations of Research 15 “Closed-ended” items, cont. Or we may count the number of symptoms We may have a cut point: e.g., moderate depression is defined as 4+ symptoms Below is a list of different feelings. Circle the number that shows how many days you felt each of these over the PAST WEEK. I was bothered by things that usually do not bother me. I felt I could not shake off the blues even with help from my friends or family. I had trouble keeping my mind on what I was doing. I felt depressed. I felt that everything I did was an effort. My sleep was restless. I was happy. I enjoyed life. I felt sad. Rarely or none of the time A Little of the Time A moderate amount of the time Most or all of the time (less than 1 day) (1 or 2 days) (3 - 4 days) (5 - 7 days) 0 1 2 3 0 1 2 3 0 1 2 3 1 1 1 1 1 2 2 2 2 2 2 3 3 3 3 3 3 # of symptoms: 0 rated 21or 3 items 0 0 0 0 0 Center for Epidemiologic Studies National Institute of Mental Health Moods & Feelings Foundations of Research Closed-ended items, summary Chief virtue: clear operationalization Specific & concrete; we know exactly what the participant is responding to Easy to quantify & use statistically Can be tested for reliability Chief liability: potential insensitivity Brief & simply worded; potentially superficial “Top down”; issues are imposed on the participant Discrimination studies: no option for “has no attitude” Attitudes / moods: not sensitive to participants’ personal perspectives (?). 16 Foundations of Research Survey formats; “Open-ended” items General textual / qualitative response; More sensitive to the respondent How have you enjoyed your methods class so far? Please list the three things that first come to mind when you think of The Foundations of Research. More difficult to interpret Can be analyzed as qualitative data (see discussion in Descriptive data.) Can be quantified; frequency counts of citations or statements “linkages” analysis of co-occurring statements Often presented as textual portrayal plus minor quantitative analysis. 17 Foundations of Research 18 Mixed survey formats Example of a mixed question format from a survey of women’s sexual practices. Closed-ended attitude scale Open-ended description Simple behavioral index. Personal Safer Sex Guidelines How strict are your personal guidelines or rules for safer sex (e.g., condom use, “safe relationships,” etc.)? 1 Not at all Strict 2 3 Somewhat Strict 4 5 6 Very Strict 7 Extremely Strict What are your rules for safer sex? Have you ever refused to have sex with someone to stay safe? 0 1 2 3 Never once or twice a few times many times Foundations of Research SUMMARY 19 Survey topics & item types Surveys assess: Knowledge Attitudes or preferences Ongoing or intended behavior Closed-ended formats Psychology 242 is a wonderful course… Highly structured, easy to analyze Potentially insensitive Open-ended formats …list the three things that first come to mind… More sensitive to the participant Potentially ambiguous or difficult to analyze Surveys typically… Use multiple items Employ several formats. Foundations of Research Survey Research Topic areas & formats ✓ General issues in Survey research Sources of bias (or fraud…) in survey research Examples of surveys 20 Foundations of Research Self-report questionnaire Forms of survey administration 21 “Paper and pencil” or internet-based; Primarily closed-ended, structured questions Limited open-ended items All data collection increasingly computer-based Assume at least moderate reading level Cheap & easy to administer Internet: Representativeness very dubious Face-to-face interview “Door step”, formal research center, or telephone Allows in-depth qualitative questions Many studies combine questionnaire & interview formats (Telephone version becoming obsolete) Foundations of Research Cost / population access Different methods are more / less likely to reach certain populations, e.g.: General issues in surveys Disfranchised / poor populations often not reached by internet or telephone Cell phones & avoidance of telemarketers less availability for telephone surveys Stigmatized populations less available for face-to-face interviews, more available via internet. Participant sophistication Participants may not be able to accurately report certain topics Attitudes toward stem cell research from readings. What factors are most important to your choice of political candidate.... Describe the amounts and types of proteins you eat during a typical week... “Rationality bias”; many questions (incorrectly?) assume a rational reason for behavior: Why do you have unsafe sex... What is your chief reason for using alcohol each night… 22 Foundations of Research Social Desirability Responding Clear face-valid items addressing embarrassing topics yield less valid responses How often are you dishonest with your friends? Have you ever cheated on an exam....? High social desirability wording elicits inaccurate responses… Do you support protecting our Nation’s forests for future generations? (Does “yes” mean you an “environmentalist”?). Do you feel there are ways your husband could be closer...? (Does “yes” make you are unhappy in your marriage”?). Populations differ in social desirability responding; may be a confound in studying group differences Women report more suicidal thoughts, but may be more willing to disclose, creating a possible confound… 23 Desirability can be minimized by: Anonymous surveys Assurances of confidentiality Computer administration (no personal interaction) Careful wording / pilot testing of items Foundations of Research 24 Social desirability responding Click image for NY Times article Do people lie on surveys? Men routinely report more sex partners than do women. If the sample is unbiased by gender, number of partners should balance for men & women. Social desirability hypothesis: Women underestimate partners Men overstate partners Much of the difference due to: A high proportion of women who report 1 partner A few men who report many partners. Possible sample bias (confound?) in who responds to such surveys? Click for article from phys.org Foundations of Research General issues in surveys: Time Frames Rare(er) events require a long time frame to assess When was your last doctor’s visit… These questions asses the last time you left a romantic relationship… Longer term recall can be surprisingly unreliable Recall of last doctor visit highly unreliable when checked against medical records Shorter time frame yields more reliable responding Memory is better for more recent effects “Exit interviews” from medical visits far more reliable than even 2week retrospective measures. Current, concrete behaviors are more accurately reported than behavioral trends. In general, how often do you miss a dose of your medication Less reliable than… Let’s go over each of the past 7 days and tell me if you took or missed your medication dose. 25 Foundations of Research General issues in surveys: Question Order Questions trigger participant’s memory or attention, and can bias questions that follow, e.g.: Do you think Social Security & Medicare payments have kept up with inflation.. then Do you favor or oppose Democratic efforts to expand Medicare payments... Bias can be limited by counterbalancing questions. Using different question orders in different versions of the survey. 26 Foundations of Research Summary 27 Survey administration Internet increasingly important as self-report method Face-to-face interviews more common in clinical research Time frames & question order can influence responses Population access & sophistication Some groups are difficult to reach Creates threat to External validity General Issues Summary Assumption that participants understand survey materials often questionable. Social desirability responding Inhibited responding threatens Internal Validity May represent a confound if groups differ in desirability set. Foundations of Research Survey Research Topic areas & formats General issues in Survey research ✓ Sources of bias (or fraud…) Examples of surveys 28 Foundations of Research Bias / Fraud in survey research Social research is increasingly important to political & cultural debates. Effects of gay marriage Political “approval” ratings Scientific consensus on global warming Research on working mothers … Pressure for confirmatory results encourages bias or outright fraud (see this week’s article on Opinion Polls). In the study structure Items used Sample In the interpretation of results “Cherry picking” Simple distortion 29 Foundations of Research Opposition to gay marriage Example 30 Example of fraudulent survey use Judges have struck down religious doctrine as a basis for disallowing gay marriage Opponents have sought evidence of civil harm to justify discrimination Harm to children has emerged as key issue in the debate APA; multiple studies show no harm. Mark Regenerus’ Survey: Children in households with gay/lesbian parent fare worse. Funded & cited widely by gay marriage opponents Study sample and interpretation of results wildly biased Has been disavowed by Renenerus’ Academic Department and April DeBoer and Jayne Rowse … after closing arguments in their challenge to Michigan’s marriage restriction. Click image for coverage. Credit Mandi Wright/Detroit Free Press, via Associated Press American Sociological Assn. Continues to be central “evidence” in court cases Foundations of Research 31 Sources of survey fraud; question wording can elicit a response desired by the researcher; Bias in survey research: Leading or biased items How much do you support the administrations’ actions to protect you and your children from terrorists… Wording can “normalize” a response, e.g., When do you feel that it is O.K. to cheat on an exam? ..when I really do not know the material .. when others are doing it .. when I think the exam is unfair Vague wording can be interpreted in a biased fashion Is there anything your husband could do to be more intimate with you? “Push” polls: a survey can be used to actually create an attitude. Foundations of Research Push Polls: An item “traps” the participant into endorsing a specific view. When the “data” are released the biased wording is ignored. 32 Publicity about the “findings” (e.g., by politically biased news organizations) are used to further create or change attitudes. Many political & social organizations use this strategy to… Ostensibly measure attitudes objectively Use the “results” to influence popular opinion. Example of a highly biased survey: The Republican National Committee ‘Future of American Health Care’ survey. The survey was distributed in several counties as part of a fund raising letter. It is clearly a “push poll” designed to create fear of health care reform. It got limited distribution, but is a great example of a Push Poll Foundations of Research GOP health survey (2009 – 2010) Some of these items are simple lies – or manipulations based on lies – designed to induce anti-health care attitudes… Other are powerful (and dishonest) emotional manipulations ✓ ✓ ✓ ✓ 33 Foundations of Research Forms of survey bias: 34 Provide leading or emotionally manipulative information to induce an attitude rather than simply measure it, to provide politically useful “data”.. Questions that, if you accept their assumptions, can only be reasonably answered one way… Foundations of Research Biased surveys: Democratic example An example from the Democrats, that is also used for fund-raising. 35 Foundations of Research Manipulative presentation of questionable information Simple emotional manipulation Distorted description that may be changed in presentation of findings “Who could disagree” item. Democratic biased survey (2007) 36 Foundations of Research Summary: Manipulating attitudes by surveys 1. Ask manipulative or highly leading questions Summary 2. Find high levels of agreement (and potentially change participants’ attitudes). 3. Publicize – and often distort or overstate – the “findings” via highly biased news sources 4. News reports themselves lead to attitude change among people who are uncertain or uninformed. 37 Foundations of Research Survey Research Topic areas & formats General issues in Survey research Sources of bias (or fraud…) ✓ Examples of surveys 38 Foundations of Research 39 Examples of surveys & data, 1 Consumer reports survey of mental health care question • Satisfaction with therapy. • Differences between types of therapy. population sample data findings Self-Identified group: Self-selected convenience sample: Attitudes & behavior Descriptive & hypothesis tests U.S. mental health care users Readers who got therapy & returned the survey, n=4000 Self-report questionnaire, cross-sectional High satisfaction for most treatments Foundations of Research 40 Examples of surveys & data, 2 “Monitoring the future” youth studies question • Social behavior • Academics • Alcohol & drug use • Health. population sample data findings Demographic group: Random sample: Knowledge, attitudes and behavior Mostly descriptive All U.S. youth, 15 -> 21 years old. Sample of High School health classes, n=3000 -> 5000. Face--to--face interviews & questionnaires, longitudinal (bi-yearly) Assess yearly trends/shifts in drugs, grades, emotional well being Foundations of Research 41 Examples of surveys & data, 3 Gallup, Time/CNN, other polls question populations samples data findings Demographic: National, random Knowledge, Attitudes, Behavior Descriptive Digit dial telephone, n=150 to >500 Brief interview, cross-sectional Ratings of politicians, Consumer preferences Approach to Affordable Care Act - Eligible voters, Political opinions, Lifestyle information Social attitudes, e.g., managed care - Target age groups Self-identified: “Democrats”… Behavioral: Voters, ACA users… General: - U.S. adult population Foundations of Research 42 Examples of surveys & data, 4 Exit polls question Election outcome, possibly stratified by state / region population sample data findings U.S. electorate Probability Self-reported behavior Descriptive/ predictive National and/or local electoral district Stratified random sample of electoral districts. Self-report interview, cross-sectional Increasingly inaccurate predictions See reading on shifts in use of polling data in U.S. politics Foundations of Research 43 Examples of surveys & data, 5 ”Social Issues Survey” of Chicago gay / lesbian community question - Stress & coping - Alcohol & drug use - Responses to HIV / AIDS population sample data findings Self-identified Targeted multiframe Attitudes & behavior Descriptive & hypotheses Self-identified gay, lesbian, & bisexual adults in Chicago. community newspapers, organizations, & mailing lists, n=3500 Self-report questionnaire, cross-sectional - High experience of discrimination - Less stress & alcohol-drug use than expected Foundations of Research 44 Examples of surveys & data, 6 National Institute on Drug Abuse Household survey of Alcohol and Drug use question Alcohol-drug use and problems, treatment use, health effects. population sample data findings National Random Multistage: Knowledge, attitudes & behavior Typically descriptive All U.S. adults 1. Census tract 2. Household, 3. Any adult member Face - to - face Interview, successive cross-sectional (each 5 years) Age & regional differences in substance use, trends over time in use & problems n>4000 Data often used for hypothesis-oriented secondary analyses (i.e., as archival data). Foundations of Research Surveys typically use multiple items to measure each hypothetical construct Summary Summary: Testing Hypotheses Correlations among items tell us if they are reliable in measuring the same construct. We use Mediating Analyses to Test hypotheses about correlations between constructs Build or test theory Cross-sectional analyses are difficult to interpret Causal direction? 3rd variable problem Longitudinal analyses help us determine causal direction 45