174AW 2007 Class notes
- 1 Main points covered in the class
- 2 Catchup for those 5 who missed the first class
- 3 A statistical analysis of ethnographic observations about the first day of class and wiki participation
- 4 Write me a short note on notebook paper in class
- 5 Course Notes
Main points covered in the class
Catchup for those 5 who missed the first class
With the confusion in starting date, fortunately, you are the only ones to miss the Thursday class yesterday, So login with your real name, use the search window to find the course site at "Human Social Complexity" then go to the class notes site and see if you can raise one of the students who used the wiki in the first class (some practiced with the five main topics of the assignments; these also outline the main structure of the course) and see if you can "talk" with them at the wiki or meet outside class, phone, etc to get a better idea of what was covered. You can also meet with the TA after class on Tuesday or email her and the others on the "catch-up" list, after going over the web pages carefully to formulate your questions. In any case, make yourself very familiar with the wiki my printing or reading on-line, get instructions from the main page about editing, login, write your own home page at your user address, etc, and follow the wiki links from our home page.
Sept 28 2007
1. Read some ethnographic cases and compare them
- What would it be like to live in one of these cultures?
- How do these two cultures differ?
2. Learn about dimensions of culture
3. What is the evidence -- systematic (SCCS)comparisons -- for these dimensions of culture
4. Sustainability comparison
5. How did we get to this situation? (Plateau in human industry)
User:Maxzibit Ta Duc?
This early experiment was a success. 35% of those present (7 missed class) made wiki entries, most with last name identity and initial, the following with full or partial lecture outlines:
- User:Aalkalby Al-Kalby
- User:Paige Austin
- User:Suzanne Nguyen
- User:Scvuong Vuong
- User:Sfaghani Faghani
And some with other responses.
- User:Jankip Janki Talk:Human Social Complexity and World Cultures
- User:Christine Black
- User:Johnreitzell Reitzell
Other notes Madamba
1. Read some ethnographic cases and compare them >1. Get a feeling for" What would it be like to live in one of these cultures?
- How do these two cultures differ? --> Read enough about different aspects of cultures to compare and contrast them.
- How you compare them depends on what you're focusing on.
2. How do these cultures differ on basic dimensions
3. What is the evidence -- systematic (SCCS) comparisons -- for these dimensions of culture
2. Learn about dimensions of culture
A statistical analysis of ethnographic observations about the first day of class and wiki participation
(Just read and be able to discuss in class on Tues Dec 2nd--its an example of using simple stats in qualitatative observational fieldwork, and illustrates techniques that can be used to test hypotheses for writing assignment #3--Doug 14:03, 28 September 2007 DRW)
As a side note, there was a unexpected human behavioral aspect to our first class meeting: We observed that those who came early sat near the door and filled the right side of the room. None of the 15 who filled the left side roster (coming in later) showed signs of participation. This can be shown as the contingency table
+ - early/late 13 16 Wiki participants 0 15 Did not participate (3 dropped the class) (fortunately, only 5 missed class: Casey, Ryan, Hsiao, Jessica, and Kay--will help get them caught up)
So let's apply some of the statistical tests we are going to use on our comparative data: First, the ordinal probability tests for a 2 x n category table, http://www.quantitativeskills.com/sisa/statistics/ord2hlp.htm. Using this on-line calculator: What is the ordinal probability that this is a random result? (K-S, for Kolmogorov-Smirnov test, is the best of these tests): __?
The Fisher Exact test, http://www.matforsk.no/ola/fisher.htm, is more powerful, here for a 2 x 2 table. What result does it give?: __? Which do you believe and why? Was this result, from observed and self-report behavior, just random? What "confidence" would you give to the statement: those coming later to class hardly participated in the wiki experiment, while those did participate (almost) all came early. Confidence rating? __? Note that this is a conditional statement and does not imply that all who came early did participate. This is already both a precise statement and one that is complex in the sense of indeterminacy. A deterministic statement would have to go further and say that all who came early did participate: but since there were 15 who did not, this is clearly wrong. Simple statistics look for determinacy (the linear model). Complex statistics allow contingency and thus incomplete determinacy ('in'-determinacy, constraint).
Note that this experiment itself was indeterminate: it was not designed or even imagined beforehand. It was a behavioral regularity observed opportunistically, after-the-fact. Most of anthropology fieldwork is of this nature: paying attention to what is going on around you, what people are 'saying' (e.g., where they sit), what they are 'doing' (e.g., whether they 'wiki'), and figuring out 'who is who' (the network part of the study).
In response to Christine's question by email: What should we do with these statistics? the answer is click open the stat calculators above and fill in the four numbers in the cells, then click calculate (if you are confused, use the second calculator and check wikipedia:Fisher's exact test. We will discuss in class.
Write me a short note on notebook paper in class
What can you learn from these observations, if you trust in these conclusions? What would lead you to trust or distrust in them? What are the two probabilities from the two different calculators above: what to they mean? Do they add to your confidence in these conclusions?
Write me a short signed note on your notepaper and let me collate the results. Lets see further if there are further contingencies between belief, trust, and behavior.
Also: do you have an hypothesis about why "We observed that those who came early sat near the door and filled the right side of the room"? That remains unexplained in these observations.
Could this have been done as an experiment? Note that a social experiment requires control, such as assigning people to sit on the left or the right side of the room, or assigning them to "come early" or "come late." Is it possible that "spontaneous" behaviors may produce (recurrent) patterns that would be difficult to reproduce experimentally? Is this another indicator of "complexity", self-organization, etc.?
Someone Did! Thanks!
Shehzad wrote a nice essay on this problem critiquing inferences that could be made -- a good job! Significance tests alone don't support arguments only the null hypothesis. As it turned out I remembered after giving these results that I had started giving instructions for the wiki 3-4 minutes before class began, i.e. the non-randomness may have been of my own making, a good example of bias in data! Good to be critical of statistics! Which reminds me of two points we can emphasize later:
Mudsticking statistical tests
- If you winnow through a bunch of tables looking for significance as your criteria for a finding, you are sure to find spurious results, a bias of your own making. Better to start with an hypothesis. If you winnow through 20 correlations, you would expect one to be spuriously significant at p < .05. This is called the mudsticking problem with significance tests: throw mud at a wall and some will stick to look like a pattern.
- There are some measures of possible sources of bias (vars 798-813) in the SCCS codebook, and it would be good to check the correlations between the single-factor scores in part 2 of the course against these variables.
- I did a test of this (17 data quality variables) for the seven factor variables we constructed last year (7x17=119 tests) and found 5 with (spurious) "signifance" < .05, almost exactly what is expected by chance. Very encouraging that perhaps these factors, composite of MANY judgments, are unbiased.
- EXCEPT that four of these at <.05 were for the STATEHOOD factor, where only 0.7 such results were expected. So maybe there is some bias in the statehood single-factor variable.
Oct 2 2007
Ethnographic reading essay
How to find the books on your societies:
- a. choose a society from this list
- b. look up society on the provided list of bibliographies to choose which books to use
- c. search for these books at the library
- d. you can use the online database, eHRAF, to search topics...but this database doesn't give you full texts! Just snippets of chapters or paragraphs,
Things to remember:
- -Write a humanistic essay about what it's like to be in the societies that you chose
- -Make qualitative judgements
- -Compare and contrast dimensions of culture
- -DO use quotes, put quoteation marks around them (this doesnt makes the information reliable but gives the source. Lots of convergent sources are more reliable than none)
- -Put in your bibliography the name of your readings (at least 2)
- -Cite your sources
- -Refer to this essay as an example of how variables are used for comparison
- -Essay should be 4 pages at most
Fisher's Exact Test
This is used to check the probability that a simple 2x2 cross-tab of two variables is due to chance. Different from the reliability of data...which is established by correlating two independent measures of the same observational variable.
Oct 4 2007
Factors of culture analytic essay
This next essay is going to explore how reliable certain dimensions of culture are based on your comparisons of the judgments made in the Codebook and your own findings. You will notice that there are some ratings that are very accurate, while some dimensions are very hard to measure.
Single factor: we're going to look for things that factor together, or all agree. This type of analysis helps with data reduction. Refer to this page for more information on how to do a single-factor test.
- "-1" = there is a negative correlation
- "0" = there is no correlation
- "+1" = there is a positive correlation
- "1" = correlation is totally random
- "0" or close to zero = e.g., p<.01 less than one chance in 100 the correlation is random.
- These are probabilities in the range 0 to 1, so there is no negative.
- Both + and - correlations could be significant, but not 0.
Oct 9 2007
After you correlate your variables, write about why you chose to construct your factor-analysis the way you did. Why did you choose the other variables that you used? Why do you think they should be paired? How are they related? Are some of the variables precursors to other ones? What do the patterns that you found imply about the societies?
Oct 11 2007
This was an intensive first tutorial following the instructions for the second paper. All or almost all students got first results.
Oct 15 2007
This was an intensive second tutorial following the instructions on the wiki. We went ever results from Thursday and students got second results.
On the 16th, Wednesday, we had a flurry of emails from students and back to the entire class email list. The PDF example was developed further, emailed, and replaced the previous iteration on the wiki. All the needed interpretations are provided, in detail. One crucial exercise is left for the student.
Oct 17 2007
Although the second paper is due we begin the 3rd exercise (cross-tabs) which may help you interpret results from the last paper. You may revise the last paper accordingly.
A new set of detailed instructions for cross-tabs and graphs is provided.
We will discuss after the tutorial how to outline and write the paper.