Anthro 174AW : Tues WEEK 4


By Tues WEEK 5: Turn in one additional page with one or two tables supporting or not supporting one of your hypotheses, and taking into account the POINTS 2 AND 3 discussed below.


1. You should now have written your proposal. Some points to note:

For each of the topics (at least two, related by your hypotheses) you need to spend some time in the library (Pauline Manaka, the social science librarian, can be of help) finding cross-cultural or other studies of these topics that can help you shape your hypotheses and formulate the introduction to your research paper. The introduction should cover why the focal topic is of interest, why you think the topics in your hypothesis are related, what you have learned from the literature on these topics.

Remember that each hypothesis is stated in terms of groups of societies:
e.g. (those groups of) societies that have specialized or full-time warriors will (correspond closely to groups of societies that) have organized political states, i.e., centralized government.

You should have an argument that precedes or directly follows the hypothesis as to why you think this will be true (what ideas or theories of society give rise to the hypothesis?).

If you find a case study that bolsters or exemplifies your argument you may discuss and show the relevance of the case to the argument.

You should test the hypothesis comparatively not with one or a few cases, however, but with either the full cross-cultural sample, working with the variables in the codebook, cross-tabulations and correlations of the variables.

You should have realized by now that you are not dealing with hypotheses about contemporary America, variation in urban societies, class or ethnic differences, but with variation across the full range of human societies -- you should familiarize yourself with what this entails in terms of the sample we are using (SCCS=STDS=standard cross cultural sample).)

Dont be reductive in your logic. For example, an argument leading up to the hypothesis "I will show that polygyny (or:X) is sexist" mixes levels: without specifying variables that are indicators or sexism (e.g., being extreme on variable Y) the statement is essentialist, e.g., you already know the answer because there is some attribute of X that is essentially sexist, etc.


2. How to build your hypothesis or hypotheses towards a research paper: RELIABILITY and definition of the categories within each variable.

For each topic, make list of the different concepts you might use in testing your hypotheses.

Under each concept, list the variables and variable numbers.

Investigation of variables that measure the SAME concept is called the study of reliability. You can correlate and cross-tab these sets of variables to see if they measure the same thing. Try to inter-correlate ALL these measures at the same time. The correlation table will have as many rows as you have variables; the same variables will appear in the columns; the ones in the diagonal of the table show each variable to be perfectly correlated with itself; and the off-diagonal correlations between variables will be symmetric.

Think about how each variable is coded (assigned category numbers: 1, 2, 3 etc). You may want to use Spss to RECODE these numbers into an order that is more useful in measuring your concept or in testing your hypotheses.

Do you have an idea why the variables differ or do not correlate perfectly if they measure roughly the same thing? Pay attention to how the different categories are defined. If necessary, consult the original studies which coded these variables to understand how they coded the variables.

Also pay attention to the amount of missing data, number of cases coded, or to categories in a variable that are NOT relevant to your hypothesis. Possibly you want to recode such categories as missing if they do not contribute to testing your hypothesis.


3. Testing Hypotheses: Measurement and Replication

So far we have only used cross-tabs and correlations without much attention to the details of what appears in the cells of the table. Now you want to fine-tune your approach. The idea is to get the CONCEPTS in your hypotheses in line with the MEASURES you use in terms of the variables, or recodings of the variables. Multiple measure of the same concept add to the reliability of the study through correlation with each other. Independent correlations of such multiple measures with measures of other concepts specified in the hypotheses are evidence for REPLICATION. You would like to show your findings use reliable measures and that the findings replicate to show consistent relationships between the concepts as stated in your hypotheses.

How you categorize the groups of cases for each variable has a great deal with how valid is the test of your hypothesis and with how you interpret the table. Some options are 1) collapse several categories of the variable into a single category that is more meaningful in terms of your hypothesis; 2) recode the categories in a different order to make the variable a more accurate measure of the concept in your hypothesis; 3) recode an irrelevant category as missing to eliminate these cases from the table.

Note: MAPTAB does cross-tabs that have a very useful feature: each row of the table is compared with other rows by a statistic called the Z-score. If the Z-score of a row is 2 or less it means that the average of this row is two or more standard deviations LESS than the average for the table in general. If the Z-score of a row is +2 or more it means that the average of this row is two or more standard deviations GREATER than the average for the table in general. These comparisons of average also use a statistic in which the distribution of cases across categories in converted to a normal distribution, so that differences between averages are meaningful. The Z-score differences of 2 or more (plus or minus) are indicators of statistically significant differences. Hence the MAPTAB tables can help you pick out findings about PARTICULAR ROWS of your table that differ from the others.

Thus, you need to figure out HOW TO USE THE CODES in each variable: how to recode variables, and exactly what you want to compare in each table.

Once you have done that, then it is often useful to instruct Spss to convert the numbers of cases in the table into percentages, making the row totals add up to 100%, for example.

Then you can rephrase and retest any given hypothesis by saying, for example: For higher values on the independent (first or row) variable, there will be greater percentages of the societies that are higher on the dependent (second or column) variable.


4. Organizing your time - working together.

Be prepared to answer questions from other members of the class or to ask questions from other members of the class regarding things you do not understand. If you can demonstrate to me that you have helped others in the class your grade will only improve. I do not grade on a curve so you will not be hurt by cooperating. The more you learn from/teach to other members of the class, the better everyone will do and the higher the grade average will be. There are plenty of things to do within the class period at the lab especially during these early weeks of the class try to expand your research project and try out the new ideas I will be giving you, ask others about them, and teach them to others if you understand them but others do not.

You are free to work in cooperative groups in class so long as your groups are permeable others may gather around and ask questions as well.