A. I will be glad to send your grade (both exam grade and final grade) to your email address. In order to do so:

(1) Please authorize me at the top of the first page of Exam 3 ("OK send grade" will do fine.) University regulations preclude my doing so without your written approval.

(2) Please make sure your preferred email address is LEGIBLE. Please print it carefully and make sure the case for each part of the address (capital letters or smalls) is correct. If you give permission but I am unable to read your email address, I will post it to your garnet account as given in Blackboard.

C. On exam times: Any graduate or undergraduate exam held during the last week of class or finals week must occur at the time scheduled in the FSU schedule of classes. Because I also teach Mondays, 3:35-6, we cannot use the first scheduled time (3:35) which would have set the exam on 12/9. Therefore we are using the next possible available time to set the exam (5:15; there is no 4:40 time), which is also when our course meets. This exam time is Wednesday 12/9 at 5:30. I cannot deviate from the FSU exam schedule so please do not ask for changes. Thank you.

For further information on FSU's policy please reference http://registrar.fsu.edu/dir_class/fall/exam_schedule.htm and scroll down past the tables to the prose.

• ON PROOF AND CAUSALITY 

• LEVELS OF MEASUREMENT

• EXAM ONE FEEDBACK

• EXAM TWO FEEDBACK

FOCUS GROUPS (on purposes and method)

SAMPLING (which you do for everything)
 

GUIDE 7: LESS STRUCTURED METHODS
GUIDE 8: ARCHIVES AND DATABASES

ASSIGNMENT 4 GUIDELINES
   (includes expository information)

ASSIGNMENT 4 FEEDBACK

ASSIGNMENT 5 GUIDELINES
(includes expository information)
(ONLINE DATABASE ASSIGNMENT)
 

THE FOLLOWING CHAPTERS WILL NOT BE ON EXAM THREE.

Wiersma, Chapter 15
McMillan, Chapters 12 and 13

These are all excellent chapters on research reports. I hope you will keep both these books and reference them in the future. However, time precludes our covering these chapters this semester.

NOTE: McMillan, chapter 11, overlaps most of the material in Wiersma, chapters 8, 9 and 10 (which is why it was not assigned). However, this chapter does make several excellent points and you may want to skim it for review.

Like many people who are not actually survey researchers, McMillan (otherwise an excellent book) confuses random selection, probability sampling, and EQUAL probability sampling.

Remember! There isnosuch technical term as "random sample".

There is a SIMPLE RANDOM SAMPLE. Its elements are selected so that each element has an equal probability of selection, and each combination of elements also has an equal probability of selection. SRS is a technical term.

Some probability samples (also a technical term) are equal probability (EPSEM) and some (such as disproportionate samples) are not.

It is important to know the difference because when you analyze the entire sample, EPSEM samples are nearly always self-weighted. Non-EPSEM samples must be weighted to provide valid estimates of the entire sample.

 
Not only FALSE but ridiculous! Causality depends on manipulated interventions, time order, logic--but not on whether your variables are numeric or categorical. After all, age in years is numeric and usually a causal variable in nonexperimental research (except for "age at death") and adult choice of religious denomination, which is nominal, is often a dependent variable.
Review causal issues HERE: 

 
FALSE. Nonsense again. It is easier to establish INTERNAL VALIDITY, or cause-effect confidence, with more structured methods but less structured methods can be equally systematic and objective. Less structured methods are useful in providing description, including systematic and objective description, "fleshing out" with insight and depth, and generating hypotheses, all of which are part of science.

 
This one is TRUE. First, because researchers using more structured methods typically have precoded responses, thereby giving them less leeway in ascertaining the response to analyze. In less structured research, investigators often create codes after the data are already collected. Second, because less structured researchers typically have more empathy in their research, which can affect human judgment. On the other hand, bias can be built into precoded responses, in terms of how questions are asked and the coded categories presented to experimental subjects or survey respondents. So, it pays for everyone to be careful.

 
This one is FALSE. Why would we expect any a priori differences in internal validity between data collected by the researcher him or herself and that collected by another party?

However, archives and databases CAN improve the EXTERNAL validity of a study. The scope in time (many archives go back decades) or place (the entire United States--or even International data--instead of the introductory psychology classroom or the Leon county school system) means that the researcher can generalize to more populations or situations than with a local, "one-shot" study. Further, many archives contain much more data on many more cases or units than a single researcher could afford to collect. This means that during analysis the researcher can institute more statistical controls without "running out of cases." Response rates may be high too (but don't forget to check all the methodological information about the database that you choose.)

Remember any problems that existed when the data were first collected are still there! This means biased questions or measures, nonprobability samples, missing data, and so forth.
 

 
Here are a few possible problems:

• Coding categories might change over time or agencies, for example, "a college degree" may be defined differently.

• If the archive goes back several years, data may be missing, especially for older years. For example, it is tougher to find tapes of television situation comedies from the 1950s than it is from the 1990s.

• If the database extends back several years, it may use archaic non-numeric codes for the data, for example: -   &   *. Computer programs such as SPSS will read these as missing data. However, back in the old key punch and "IBM-card" days, an ampersand (&) might represent the number 12. (This means you may originally have to read your data in as alphabetic, recode it, then change the data type to numeric.)

• Similarly, the database may use alphabetic codes, such as "male" or "female". SPSS reads these as "string variables" and allows only very limited analysis with them.

 
Structured surveys are standardized. They contain a lot of questions with precoded answers. All questions must be asked in the same order, hence STRUCTURE.

• They are helpful when you want to survey many, relatively well-developed topics relatively quickly.
• If your survey uses probability sampling, it means that you can generalize the results back to the population.
• The results are also relatively quick to code. If you use CATI, your data may even be "SPSS-ready" when you leave the field.

Focus groups are relatively unstructured. Neither the number, the order, or even which questions are used is standardized (typically).

• They allow you to explore a smaller number of topics in depth.
• They allow interaction among group members and thus can generate new ideas and ideas for policy.
• You have a variety of viewpoints.
• You can try out persuasion campaigns and get some idea of how well they might work.

 
Diversity on the SAME TOPIC, absolutely! Some of the things that you hope to accomplish with a focus group are getting group discussion from a variety of viewpoints and generating possible policy solutions, again from a variety of perspectives.

Note that the desire to have a variety of viewpoints may provide an argument against EPSEM probability sampling, unless you are assured that such diversity exists in the population (for example, the overwhelming majority of parents want some kind of sex education in the schools, so you would not be able to produce a lot of variety in a focus group on this topic unless you either employed a disproportionate stratified sample, or went to some form of quota or purposive sampling.)
 

 
Typically, the field is very rich. If you do not enter with a working design and some focus, you may be so overwhelmed that you won't be able to make meaningful observations.

Objectivity may be a problem. It is a constant delicate balancing act between empathy and objectivity.

Access! You are a guest in the study locale. A lot of organizational representatives don't want to let you in. Those that do may be very atypical.

If you take a probability sample "of one," there is a good chance of getting an "extreme case." However, you are very limited in generalizing from a purposive or quota sample.

You will probably need some help from an assistant to ensure that your observations are reliable, objective, and comprehensive.

An assistant may be able to gain entrance where you would be obtrusive--or illegal (bathrooms, for example.)
 

• Access to historical archives

• Being able to understand the language or dialect used in documents

• Completeness of data available

• Documents were biased during their creation

 
Molecular behavior (smiles, waves) are more objective and easy to code.

Try some holistic subjective measures too (keeps to self) but see if you can define what you mean by them, give some behavioral examples, and try a second observer.

Can you measure duration? Frequency? How about intensity? That is more "internal" so you may have to be ingenious.
 

 
HINTS: Check out any possible alternative causal explanations. How about history? In response to civil rights activism, the following two events may occur: more Black coaches may be hired and legislation is passed. Thus, the apparent causality may be spurious--i.e., an accidental byproduct of the original causal variable (civil rights activism.) Are there any comparison groups that are comparable, such as other kinds of college or universities?

 
Here are a few suggestions:

• Start by being in the field at least a few weeks before you start formally collecting data. This way both kids and teachers can get used to you. 
• Don't take notes in front of staff or kids (participant observers spend a lot of time taking notes in bathrooms) if possible. 
• Wear clothes WITH POCKETS so you can stick notes in them.
• See if there are some simple chores you can do while you observe (e.g., put materials on a bulletin board or clean up after an activity) so you "blend in" to your surroundings.

 
Remember that researchers doing less structured research design may be in the field for months or even years.

It takes time to describe a culture or subculture. 

Further, if the researcher wants to assess the field from the viewpoint of the participants, it will take time to play different roles (remember Governor/Senator Bob Graham, who did--and still does--several "work days" each year in different jobs to see how the "average worker" sees life. Senator Graham comes from an extremely wealthy family and this is one way he has developed empathy with his constituency.)
 

 
A is only one measure (that we typically don't use in ethnographies); B is just nonsense (Solomon 4 group designs DO use repeated measures for half the groups); D speaks to reliability instead; and only B uses multiple methods in the field situation.

 
First choice would be (C), an oral history. You probably want to know about what led up to these faculty decisions and what their early experiences were. These will be way too long for most focus groups plus some information may be sensitive or emotional and the person will be less embarrassed if a group is not present.

 
An oral history will be too long for this topic and privacy probably isn't needed. Diverse ideas may make for a lively discussion so a (A) focus group will work well here. 

 
Although a structured survey is the most direct route, (D) gleaning can be useful here. Take a look at: political cartoons on doors, lapel buttons for a favored candidate, or bumper stickers in faculty and staff parking lots. 

 
FALSE. Well, they may be subject to fewer of the current researcher's biases, but it's a safe bet the original researchers had plenty of biases of their own. For example, the original coding of the textual responses to "what do you look up on the Internet" in the Public Understanding of Science data never mentions eBay, entertainment (EXCEPT SPORTS), or love and relationships. Remember those suicides called "heart failure" on death certificates to avoid embarrassing the survivors too.

 
ABSOLUTELY TRUE. There may be verbal instead of numeric codes ("M" and "F" for gender) and "wild characters" such as dashes or ampersands (&). These were read by the very old IBM or Hollereith punch card readers. However, modern computer statistical programs often don't know what to do with these types of codes and allow only limited analyses with them. Recode to numeric before starting your analyses.

 
FALSE. Any problems that existed when the original research was done, such as double-barrelled items or getting household information from 15 year olds is still there when the data are reanalyzed. Find out as much as you can. Many archives have the questionnaires, interviewing methods, or sampling information in separate sites on the Internet. Download, print, and examine these before you begin to work with the data.
 

 

VERY HELPFUL HINT (AGAIN)

BE COMPLETE!

SEVERAL STUDENTS LOST CREDIT ON EXAMS AND ASSIGNMENTS BECAUSE ANSWERS TO SOME SECTIONS WERE NOT COMPLETED.

PLEASE LOOK OVER YOUR EXAM CAREFULLY BEFORE YOU TURN IT IN. WE WILL BE GLAD TO DOUBLE CHECK FOR COMPLETENESS FOR YOU AS LONG AS YOU UNDERSTAND NO GUARANTEES ARE MADE ABOUT THE ACCURACY OF THE ANSWERS!