Overall, the assignment is to investigate a cultural domain. You need to find out what the elements are, and how they are perceived in relation to each other, and what underlying dimensions people use to distinguish among the elements.
The assignment is to obtain freelists from 100 or more respondents. I strongly advise working in teams so that the data collection is not too burdensome.
Collect the data from 100 or more respondents. Use your friends, use strangers, whatever. A random sample is not required for this assignment. For each person, give them a short set of instructions like:
"Please list all the kinds of jobs at XYZ Manufacturing that you can think of. Don't take more than 3 minutes to do this. Thanks for helping out!"
For certain domains, people are unfortunately used to being asked about themselves rather than the domain, so you need special instructions to counter that. For example, to ask about reasons that people give for quitting a job, you might say something like:
"Please list all the reasons you have heard people give for quitting a job (any job). This is not about which reasons you personally have had for quitting a job, nor about reasons that you personally consider legitimate. Rather, this is about listing all the reasons that you have heard others give. Thanks for your help!"
After obtaining frequencies, the next thing to do is to think about the list and ask yourself what the results reveal about the belief systems of the informants. In class, we analyzed the "causes of breast cancer data". When we looked at the freelists of the Salvadorean and Mexican women, we had a few interesting ideas about the most frequent mentions. One ideas was a kind of moralism: they seemed to be thinking that breast cancer is caused by "being bad", as in having sex, smoking, drinking, and being "wild". Another interesting idea was that, because cancer is hidden, they confuse cause and effect. They see bruises, and they problems with giving milk, and then later the cancer is revealed, so they think the bruises and milk problems caused the cancer, but it might be that these things were just symptoms of the still-hidden cancer.
You also need to analyze the respondent-level statistics. Correlate them with the demographic or other data that you collected about each respondent.
The assignment is to obtain pilesort data from 50 or more respondents, then run MDS on the aggregate proximity matrix, interpreting the results as best you can.
1. Select the most frequent items from the domain you freelisted. Don't use items mentioned by only one person. Arrange the items in alphabetical order and number them from 1 to N.
2. Create several decks of cards containing the item names and numbers (on the back).
3. Get 50 or more respondents to sort the cards. This is time consuming! You must act quickly!
4. Enter the data in the computer. The format will be given elsewhere. An example dataset is available here.
Step1. Number all the items in the domain in alphabetical order, like this:
Step 2. Write the name of each item in the domain on a separate card and then write the number of the item on the back of the card. I use 3-by-5 cards cut in half if I do this manually, otherwise I have the computer create a set of cards. It is generally helpful to make many different decks so that you can collect data from several people at the same time.
Step 3. Get 50 respondents to do the pilesort task. This is what you tell them:
On each of these cards, I've written the name of a [fill in your domain here, e.g. "vegetable", "reason for quitting a job"]. Please sort the cards into piles according to how similar they are. You can use as many or as few piles as you want. If there is any card you don't understand, just put it aside and sort the rest of them. Go!
After they have finished sorting into piles, ask them about the sort. One way to get them talking is to ask them for a name for each pile. You can also ask them about the characteristics that led them to put items in a given pile, as in "what do all of these have in common?". Record this information for use later in helping to interpret the MDS picture.
To record the pile information, take each pile, turn the cards over, and record the numbers like this:
Respondent's name: _________________
Pile 1: 16, 3, 12, 5 "all the really sweet fruit"
Pile 2: 1, 4, 2, 9, 8 "the bitter stuff"
Pile 3: 10, 11 "stuff you make pies from"
etc.
Make up transparencies with the key information and present the results of your study in class. The slides don't have to be beautiful, but the from an intellectual point of view you should be presenting a finished product -- a well-thought-out set of conclusions is a must.
The written report is due anytime up to the last day of class. If you submit it earlier in the semester, I will grade it and return it within a week, and then if you don't like the grade, you can make changes and resubmit it, and I will regrade it.