1. Research Methods in Social Relations Professor Mike Gallivan Georgia State University Atlanta, Georgia, USA Class 5: June 22, 2009

2. Overview of Class 5 today  Remind students about tomorrow’s assigned reading  We will discuss some material that I did not have time to present last Thursday / Friday  Different modes of data collection  Scale uni-dimensionality vs. multi-dimensionality  New material for today:  Multiple regression analysis  select only certain groups  “dummy coding” for nominal variables  Different types of measurement scales  summative (Likert) scales are most popular  also Differential, Guttman and Semantic differential  Explain how qualitative research process differs

3. Date Readings in 6 th ed. red book Readings in 7 th ed. green book Special comments June 22 7: 146-169 (omit 7: pp. 159-162) 7: 158-177 (omit 7: pp. 167-170) Finish data analysis team project. Focus on Likert scales (Semantic, differential and Guttman scales are used less) June 23 6: 128-142 article about response rate on website You can omit: p. 360 (formulas) p. 369-370 (Adv. CFA and SEM) 8: 181-195 article about response rates on website. You can omit: p. 360 (formulas) p. 369-370 (Adv. CFA and SEM) Paper from JAIS by Sivo, Saunders et al. (2006). Discuss issues related to response rate. Student will read and critique a study about its response rate June 24 19: 463-476 19: 457-460 (repeat) 20: 521-536 19: 483-491 Writing academic research papers (skip section about meta- analysis in 7 th ed.)

4. Questions you asked about U.S.A.  What is your favorite NBA sports team?  My hometown is Boston  So, my favorite team is Celtics!  Are all American people good at dancing?  No, definitely not!  Many Americans are very bad dancers!

5. Review of last week’s classes  Please take 1 minute to write one idea or lesson you learned Thursday or Friday that you will remember when course is finished!

6. Review of Discriminant Validity in SPSS  Exploratory factor analysis (EFA)  This shows which items go together  In this method, items that load onto the same factor can prove convergent validity  In this method, items that load onto separate factors can prove discriminant validity  Convergent + discriminant validity = construct validity  Performing EFA using SPSS  Procedure is called “Data reduction – Factor”  Results are easier to understand if you do:  select “Options” and then “sorted by size”  also “suppress values with absolute value < 0.40  also select Rotation: “Varimax” or “Direct oblimin”  difficult to interpret if construct is multi-dimensional

7. Review of Multi-dimensionality  Multi-dimensional vs. uni-dimensional  Some constructs have > 1 dimension  Family’s socio-economic status  Amount of salary income (from formal job)  Amount of investment income (from bank)  Amount of other income (from gambling, eBay)  Level of parents’ education  Individual creative style construct  Originality (number of original ideas generated)  Rule-conformity (willing to follow rules)  Efficiency (perform tasks in time-efficient way)  Multi-dimensional constructs in data set  Section #9 – Psychological and physical strain  Items about psychological strain factor separately

8. Topics we did not cover in Class 3 Summary of Contents from Ch. 10 (Ch. 5 in green book, pp. 96-110)  Modes of direct questioning  Paper-and-pencil questionnaire  Face-to-face (in-person) interviews  Telephone interviews  Web-based surveys on the Internet  Experience sampling (“diary methods”)  Other modes of measurement (indirect)  Collateral data (another person also gives data)  Observing others (it may be unobtrusive)  We will not discuss these methods in detail

9. Topics we did not cover in Class 3 Summary of Contents from Ch. 10 (Ch. 5 in green book, pp. 96-110)  Paper-and-pencil questionnaire  Advantages  Authors claim that paper surveys are inexpensive  Minimize interviewer (researcher) bias  Less time pressure for subjects  Subject has feeling of being anonymous  Disadvantages  Subjects cannot ask a question if they are confused  Often a very low response rate (e.g., < 10%)  You may not know how non-respondents are similar to or different from the survey respondents  Perhaps the respondents are 60% female, 40% male, but the non-respondents are 30% female, 70% male

10. Topics we did not cover in Class 3 Summary of Contents from Ch. 10 (Chapter 5 in green book)  Face-to-face (in-person) interviews  Advantages  Can “probe” the subject further for more information  Subject can ask the researcher for clarification  This may create a higher quality of information  Can provide visual aids (maps, photos, examples)  Disadvantages  Potential for large interviewer effect (bias)  Especially for sensitive or personal questions  May not want to give an honest answer to question “Have you ever had an unwanted pregnancy?” This is called social desirability bias  Some subjects might want to impress the interviewer “How much alcohol do you normally drink in 1 week?”  Both of these effects might hurt the information quality

11. Data collection modes Topics we did not cover in Class 3 Summary of Contents from Ch. 10 (Chapter 5 in green book)  Telephone interviews  Advantages  Combines advantages of both surveys + interviews  Can probe subjects for further information  Cost is low, compared to face-to-face interviews  Disadvantages  More expensive than paper-and-pencil surveys  Chance for response bias People with phones might differ from those without Famous example in American presidency in 1960s Especially for sensitive or personal questions  If you rely on a telephone directory In U.S., telephone directories do not list mobile phones Young people only have mobile phones (no land phone) So, you might have an age bias effect in your results The solution to this problem is random-digit dialing

12. Data collection modes Topics we did not cover in Class 3 Summary of Contents from Ch. 10  Web-based surveys on Internet  Advantages  1: Avoid the problem of transcript errors  2: Very, very cheap to do  3 Easy to conduct  4: No effort required to mail survey back  5: More voluntary than a face-to-face  6 Faster, easier and “more fun” for subjects  Disadvantages  Response rates are often very low (e.g., <1%)  How many subjects did you send survey to?  This information is required to compute response rate  Not a good idea to just post survey to a website unless you know how many total people will see the survey  Experience sampling (“diary methods”)  Can use a pager to remind subjects to record data  These methods can be used in medical research

13. Summary of Contents from Chapter 3: pp: 53-66 (Ch. 4 in green book, pp. 85-90 and Ch. 7: 152-158)  Single-item vs. multiple item scales  Single-item is acceptable sometimes  Some examples of single-item measures  How old are you in years? Do we need to also ask “how old are you in months?”  What is your sex? (male or female)  What is your major?  What is the city where you were born?  Advantages to multiple-item scales  multiple items required to calculate internal reliability  journal reviewers prefer multiple-item scales  you can compute amount of measurement error  measurement error = 1 – Cronbach a

14. Summary of Contents from Chapter 7: pp: 146-169 (Ch. 7 in green book, pp. 158-177)  Different types of rating scales  Graphic rating scales  Itemized rating scales  Comparative rating scales  Source of the ratings  Self-ratings (individual)  Often the individual is best judge of their own opinions  Parent ratings of child’s behavior  This is common in medical surveys for young children  Peer or co-worker or supervisor ratings  Sometimes more accurate data from another person

15. Summary of Contents from Chapter 7: pp: 146-169 (Ch. 7 in green book, pp. 158-177  Problems with all measurement  Social desirability bias – try to give answer subject thinks the interviewer wants to receive  “Do you ever drink > 3 drinks at dinner or party?”  “Have you had premarital sex?”  The direction of bias effect might differ depending:  If the researcher is a peer member (e.g., same age)  If the researcher is much older or a medical doctor  Sometimes the wording creates bias  My doctor asked me last month:  “You always wear sunscreen, don’t you?”  What would be a better way to ask question?

16. Summary of Contents from Chapter 7: pp: 146-169 (Ch. 7 in green book, pp. 158-177  Remember definition of measurement in Class 2:  Any measurement item is comprised of 3 parts:  The construct of interest (general intelligence)  Other constructs not of interest (English language skill)  Random measurement error (mistakes in answer)  Observed score (e.g., each survey item) =  True score + systematic error + random error  Variable of interest + variable not of interest + random error  Multiple item scales help calculate random error  Halo bias (generosity error) is a systematic error  Example of personal attractiveness (an attractive person is often considered smarter, more competent, trusted)  Example of website attractiveness (a website that is more visually attractive is often trusted more by users)

17. Summary of Contents from Chapter 7: pp: 146-169 (Ch. 7 in green book, pp. 158-177  Developing multiple-item scales  Domain sampling  start with very broad range of items  have some positive-worded items  “I like Presidential candidate Obama”  also have some negative-worded items  “I don’t trust Presidential candidate Obama”  Process of pilot testing the survey  Small sample to identify wording problems  Modify or delete confusing items  Then test with a much larger sample of subjects

18. Example of Survey Development: Personal Innovativeness with IT  Initial survey items:

19. Measuring multi-item scales  Examine Cronbach a (internal reliability)  Delete any items that reduce a < 0.70  For all items retained in the final scale:  Average the items together  Examine Exploratory Factor Analysis  This is a more complex method  Some researchers prefer this method of “factor scores”  Be sure to select “Options”  Missing values: exclude cases “pairwise”  “Suppress absolute values < 0.40”  Then choose “Scores” and “Save as Variables”  This allows each item to have a different weight  (e.g., item with 0.84 factor loading will have twice as much weight as another item with 0.42 factor loading)

20. Example of multi-item scales using Psychological & Physical Strain items  The next 2 slides show items from the survey

21. Section #9: During the past week… 9.During the past week, how often did you experience the following: (Circle one number per item) 1.I was bothered by things that usually don't bother me. 2.I did not feel like eating; my appetite was poor. [1] Rarely or None of the Time (Less than 1 Day) [2] Some of the Time (1-2 Days) [3] Occasionally or a Moderate Amount of Time (3-4 Days) [4] Most or All of the Time (5-7 Days) 3. I felt that I could not shake off the blues even with help from my family or friends. [1] Rarely or None of the Time (Less than 1 Day) [2] Some of the Time (1-2 Days) [3] Occasionally or a Moderate Amount of Time (3-4 Days) [4] Most or All of the Time (5-7 Days) 4.I had trouble keeping my mind on what I was doing. [1] Rarely or None of the Time (Less than 1 Day) [2] Some of the Time (1-2 Days) [3] Occasionally or a Moderate Amount of Time (3-4 Days) [4] Most or All of the Time (5-7 Days) [1] Rarely or None of the Time (Less than 1 Day) [2] Some of the Time (1-2 Days) [3] Occasionally or a Moderate Amount of Time (3-4 Days) [4] Most or All of the Time (5-7 Days) [1] Rarely or None of the Time (Less than 1 Day) [2] Some of the Time (1-2 Days) [3] Occasionally or a Moderate Amount of Time (3-4 Days) [4] Most or All of the Time (5-7 Days) 5. I felt depressed. [1] Rarely or None of the Time (Less than 1 Day) [2] Some of the Time (1-2 Days) [3] Occasionally or a Moderate Amount of Time (3-4 Days) [4] Most or All of the Time (5-7 Days) 6. I felt I that everything I did was an effort.

22. 7.I felt fearful. [1] Rarely or None of the Time (Less than 1 Day) [2] Some of the Time (1-2 Days) [3] Occasionally or a Moderate Amount of Time (3-4 Days) [4] Most or All of the Time (5-7 Days) 8.My sleep was restless. [1] Rarely or None of the Time (Less than 1 Day) [2] Some of the Time (1-2 Days) [3] Occasionally or a Moderate Amount of Time (3-4 Days) [4] Most or All of the Time (5-7 Days) 9.I talked less than usual. [1] Rarely or None of the Time (Less than 1 Day) [2] Some of the Time (1-2 Days) [3] Occasionally or a Moderate Amount of Time (3-4 Days) [4] Most or All of the Time (5-7 Days) 10.I felt lonely. [1] Rarely or None of the Time (Less than 1 Day) [2] Some of the Time (1-2 Days) [3] Occasionally or a Moderate Amount of Time (3-4 Days) [4] Most or All of the Time (5-7 Days) 11.I had crying spells. [1] Rarely or None of the Time (Less than 1 Day) [2] Some of the Time (1-2 Days) [3] Occasionally or a Moderate Amount of Time (3-4 Days) [4] Most or All of the Time (5-7 Days) 12.I felt sad. 13.I could not get "going". [1] Rarely or None of the Time (Less than 1 Day) [2] Some of the Time (1-2 Days) [3] Occasionally or a Moderate Amount of Time (3-4 Days) [4] Most or All of the Time (5-7 Days) [1] Rarely or None of the Time (Less than 1 Day) [2] Some of the Time (1-2 Days) [3] Occasionally or a Moderate Amount of Time (3-4 Days) [4] Most or All of the Time (5-7 Days) 13. I could not “get going.” [1] Rarely or None of the Time (Less than 1 Day) [2] Some of the Time (1-2 Days) [3] Occasionally or a Moderate Amount of Time (3-4 Days) [4] Most or All of the Time (5-7 Days) 8. My sleep was restless.

23. Results from Factor Analysis This is first result table showing the number of factors Initial Eigen- values Compo- nent Total% of Variance Cumulative % 15.931 45.627 21.087 8.35953.986 31.026 7.89261.878 4.9056.96368.841 5.7035.40574.246 6.6765.20079.447

24. Exploratory Factor Analysis results This is a very messy 3-factor result Most items load on Factor 1 and 2 A few items load separately on Factor 3 Total Variance Explained Initial Eigenvalues Component Total% of Variance Cumulati ve % 1 5.93145.627 2 1.0878.35953.9863 1.0267.89261.8784.9056.96368.8415.7035.40574.2466.6765.20079.447 Extraction Method: Principal Component Analysis. Total Variance Explained Initial Eigenvalues Component Total% of Variance Cumulati ve % 1 5.93145.627 2 1.0878.35953.9863 1.0267.89261.8784.9056.96368.8415.7035.40574.2466.6765.20079.447 Extraction Method: Principal Component Analysis. Total Variance Explained Initial Eigenvalues Component Total% of Variance Cumulati ve % 1 5.93145.627 2 1.0878.35953.9863 1.0267.89261.8784.9056.96368.8415.7035.40574.2466.6765.20079.447 Extraction Method: Principal Component Analysis. Total Variance Explained Initial Eigenvalues Component Total% of Variance Cumulati ve % 1 5.93145.627 2 1.0878.35953.9863 1.0267.89261.8784.9056.96368.8415.7035.40574.2466.6765.20079.447 Extraction Method: Principal Component Analysis.

25. Review Data Analysis Assignment  Which surveys items are Nominal?  Other names for “nominal” items  Categorical or dichotomous (if just 2 categories)  Chi-squared (“crosstabs”) analysis that Dr. Doug Rice had you do Friday was a dichotomous item  Items? (section # and item #)  Which surveys items are Ordinal?  Items? (section # and item #)