Previous - Next
Day 1: Opening comments and introductions.
Discussion of three topic areas: domain of women's health and illness, context of women's health, and women's health-seeking behaviors.Review of the agenda.
General introduction to qualitative methods.
Day 2: Unit 1 - Direct observation.
Unit 2 - Key informant interviews.Key informant interviewing (Do role play: demonstration).
Conducting observations.
Managing textual data.
Overview of other methods: Pile sorts, free listing, paired comparisons, ranking, focus groups.
Day 3: Unit 3 - Free listing.
Day 4: Fieldwork day practice - Units 1-3.
Day 5: Unit 4 - Free listing - Unit 5 free pile sort.
Day 6: Fieldwork day practice - Units 4-5.
Day 7: Unit 6 - Key informant interviews.
Day 8: Unit 7 - Paired comparisons.
Days 9-10: Fieldwork days practice - Units 6-7.
Day 11: Unit 8 - Preliminary analysis and developing the coding system.
Day 12: Review and planning for future units.
Each unit should include a piece of didactic presentation, demonstration of the technique(s), role playing, practice in the classroom, and field practice.
The process for entry and analysis of pile sort data involves the following steps:
1. Enter the Anthropac subdirectory and load the program:cd anthro32
anthro2. Enter the editor function of the anthropac program (to enter pile sort data):
select EDIT
3. Give the data file a name, such as "PSWOMEN1.DAT" (standing for pile sorts of women, first dataset)
4. Enter the data in what is known as the Boster format, where each line of data indicates a different pile, and the first number in each line is the informant id number. For example, here is a dataset with two respondents and ten different items:
1 2 4 6 8
1 1 3
1 5 9 10
These data lines indicate that there were two respondents, number
1 7
1 and number 2. Number 1 made four piles, with items 2,4,6,8 in
2 4 6 8
the same pile, 1 and 3 in a separate pile, 5,9,10 in a third pile and
2 1 3
item 7 in a pile by itself. Respondent number 2 made 5 piles, and
2 2
sorted things a bit differently.
2 5 9 10
2 7
F2 to save the file
ESC to exit from the EDITOR5. Do an initial analysis of the data.
Go to COLLECTION
Select PILE SORTS
Select SINGLE/FREE
Select BOSTER
Enter name of datafile - ie, PSWOMEN1.DAT6. Do a Multidimensional Scaling (MDS) analysis of the data.
Go to ANALYSIS Select SCALING Select MDS Enter the name of the aggregate proxmities data file (de, AGPROX.DAT)
The MDS diagram that appears will have a stress score associated with it. A stress score of less than 0.15 is considered acceptable.
The diagram itself has numbers which indicate the relative placement of specific items in the diagram.
7. Create a set of labels for the women's illnesses:
Go to EDITOR
Create a new file for the labels, with a name like
PSWOMEN1.LABEnter the labels sequentially in the file:
Khamjoori
Safed Paani
Pet Dukhe
Kamar Dukhe
Ochu Masik
Vadhare Masik
Anyamit Masik
Taav
Khansi
Rataundi
F2 to save
ESC to exit8. Generate a labeled version of the MDS.
Go to MANAGEMENT
Select PLOT
Select SCATTERPLOTEnter COORD.DAT in space for Coordinates file, then enter to move to the next screen
Enter PSWOMEN1.LAB into space for File contain point labels, if any.
The resulting MDS picture will have each item labeled in the diagram.
9. To print out the MDS, or any other output:
Go to OPTIONS Select OUTPUT DESTINATION Select PRINTER to send all output from that point on directly to the printer or,
Select FILE to send all output from that point on to a computer file that you specify. This file can then be edited using a word processing program.
Purpose
The Triad sort technique can be used for the same purpose as pile sorts - to find out which categories of items (e.g., illnesses) are seen to "go together," and which are separate or "different." This unit may be used if the interviewers are unable for some reasons (e.g., With illiterate informants) to perform pile sorting.
Advantages and Disadvantages
Advantages: The triad sorting task can be used with illiterate persons. The interviewer reads off each triplet of items and asks the respondent "Which of these three is the most different from the others?" or "Which of these three doesn't fit with the others?"
Disadvantage: It is necessary to have the ANTHROPAC computer program to design the triads interviews, and to analyze them. A large number of triads are required to be presented to respondents. Many will become bored with the process.
Data Collection: (see the flow chart)
1. Identify the topical domain to be explored; select a list of items (usually no more then 15 to 18 items).
2. Enter Anthropac and make a labels file for common women's illnesses.
Using ANTHROPAC, select <EDIT>. Make a label file of women's illnesses "WOMENILL.LAB" by writing the list of items, with a carriage return after each item. The label file will look like this:Khamjoori
Safed Paani
Pet Dukhe
Kamar Dukhe
Ochu Masik
Vadhare Masik
Anyamit Masik
Taav
Khansi
Rataundi
F2 to save the file
ESC to exit from EDITOR
3. Generate data collection forms:
Go to COLLECTION
Select TRIADS
Select MAKE
Select BIBD
BIBD stands for "balanced incomplete block design; it is a way or minimizing the number or triads created without sacrificing comparability.
Select the appropriate number of items (corresponding to the label file). If there are 10 illnesses in the file, then select < 10> (Select by moving the arrow key down).
Next select the size of the total list of triads. Notice that for 10 items, you have a choice of a "Lambda 2" design - 30 different triads or a "Lambda 4" design - 60 different triads. Sixty is a lot of triads and can become boring for the informant and interviewer; so choose the <Lambda 2> because it has only 30 triads.
The next menu asks for the name of the label file. Type <illness.lab>. "Which questionnaires to make?" is asking how many respondents are to be interviewed. If the plan is to interview 20 informants, then make about 22 interviews (it is good to have a couple of extra interviews).
WRITE: 1-22 answer the question.
"Randomize order of triads..." <Yes>
"Randomize order of labels... <No>
"Respondent instruction file" <TRIADSQ.INS> (this file can be modified to suit local needs).
Note: The program now creates a fairly big file. It consists of all, as indicated, is: N15L1D1R.QST.
4. Print triads data collection forms:
Go to COLLECTION
Select TRIADS
PRINT Be sure there is enough paper in the printer. Print the set of questionnaires.
Note that each questionnaire will have a unique id number at the top of the form.
5. Go out to the field to collect the 20 interviews (read off each triad and ask the informant which item does not belong to that group).
6. Enter triads data onto the computer:
Go to EDITOR
Give the file a name like: TRWOMEN1.DAT
Enter each form's data all in a line, with the id number on each form as the first character. The remaining numbers should be 1, 2, or 3 depending on which column is circled. For example,
8123232312111323123121231231232
9232112332123112332211233312223
The above data indicate that two forms were entered, one for id number 8 and one for id number 9. Each respondent completed 30 different triads. For respondent number 8, the first item circled was the item in column number one, triad one; the second triad had the item in column number 2 circled, etc.
F2 to save the file
ESC to exit from EDITOR
7. Unscrambling the triads data. The triads data entered in step 6 need to be unrandomized:
Go to COLLECTION Select TRIADS Select SCORE Select BIBD (if appropriate)
Identify the number of items (i.e., 10 items) Identify the Lambda used (i.e., Lambda 2)
Enter the name of the file containing the raw triads data
Respond "Yes" to "Whether questionnaires were randomized?"
Respond "No" to "Different design used for each respondent?"
8. Do a Multidimensional Scaling (MDS) analysis of the data:
Go to ANALYSIS
Select SCALING
Select MDS
Enter the name of the aggregate proximities data file (i.e., AGPROX.DAT)
The MDS diagram that appears will have a stress score associated with it. A stress score of less than 0.15 is considered acceptable.
The diagram itself has numbers which indicate the relative placement of specific items in the diagram.
9. Create a set of labels for the women's illnesses:
Go to EDITOR
Create a new file for the labels, with a name like PSWOMEN1.LAB
Enter the labels sequentially in the file: Khamjoori Safed Paani Pet Dukhe Kamar Dukhe Ochu Masik Vadhare Masik Anyamit Masik Taav Khansi Rataundi
F2 to save
ESC to exit
10. Generate a labeled version of the MDS:
Go to MANAGEMENT
Select PLOT
Select SCATTERPLOT
Enter COORD.DAT in space for Coordinates file, then enter to move to the next screen
Enter PSWOMEN1.LAB into space for File contain point labels, if any.
The resulting MDS picture will have each item labeled in the diagram.
11. To print out the MDS, or any other output:
Go to OPTIONS
Select OUTPUT DESTINATION
Select PRINTER to send all output from that point on directly to the printer or,
Select FILE to send all output from that point on to a computer file that you specify. This file can then be edited using a word processing program.
Purpose
This appendix presents a method for entering and managing textual data using a microcomputer. The data from every exercise described in this protocol should be entered onto a computer, particularly those involving large amounts of textual field notes. The database created with this exercise will form the core for a qualitative database which will grow with time as the organization completes more research.
Instructions
1. The solution proposed here for managing textual data is the use of a word processing software, such as WordPerfect 5.1, combined with a text retrieval and management software program called ZyIndex or DtSearch1. Henceforth, I will refer to DtSearch only, meaning DtSearch or ZyIndex.
1 The author has reviewed several other textual management programs (e.g., Ask Sam, Word Cruncher, DtSearch, etc.), but has found DtSearch optimal for the purposes described.)
2. The complete process for managing, coding, and retrieving qualitative data is outlined in figure 1. Raw field notes (i.e., handwritten) should be written as expanded field notes on a microcomputer, using a word processing program such as WordPerfect. The notes should be translated into English, leaving illness and other special terms written in the English form of the local language. Alternatively, all terms can be translated. But it is important to include the original term in parentheses, because often a translation does not convey the shades of meanings contained in the original term.3. Careful attention should be paid to consistent spelling of local terms.
4. Each data file should have a unique, meaningful name. Each file name should represent a discrete data collection unit: for example, one interview, an observation of a key event, a focus group session. Several projects in India are using a format in which the first two characters indicate the type of data collected (i.e., key informant interview, focus groups, etc.) the next six indicate the date in year, month, day order; the first two characters of the file extension are the initials of the data collector and the last digit is an order number for the day (for example, KI900711.JGI would be the file for the first key informant interview conducted by Joel Gittelsohn on July 11, 1990).
5. The field notes should be kept on a separate subdirectory of the hard disk (preferably) or on floppy disks if the computer has no hard disk.
6. Install DtSearch on the computer, and index the qualitative data files. Common words like articles and conjunctions, called noise words, are ignored. DtSearch manages existing data files without altering them, requires no keywords, is equally useful for archived files and for files actively in use, and can recognize a wide variety of word processing formats, plus 1-2-3, DBASE III, and ASCII. DtSearch's second program function, searching, uses index lists to find whatever is specified. Searches may be refined by the use of Boolean operators, wildcard characters, parentheses and a "search within" feature that lets the searcher specify the number of words apart two items can be. Located text can be marked and extracted. Unlike some programs that pull out isolated "hits," DtSearch shows the full text, permitting the user to see the context of the "hits." The program satisfies the majority of criteria for textual data management software.
Purpose
The exercise will help in devising a system for coding qualitative data using microcomputers. Coding the qualitative data will help when trying to locate needed materials in a large database.
Even after field notes have been entered on a word processing program and indexed, it may still be difficult to find certain kinds of information. For instance, a direct observation of a woman ordering her husband to buy her a dress might be a good example of household decision-making in action, but how would it be located using DtSearch? The solution lies in the use of codes.
What are codes? Miles and Huberman define a code as "an abbreviation or symbol applied to a segment of words...in order to classify the words... they usually derive from research questions, hypotheses, key concepts or important themes" (Miles and Huberman, 1994). Coding "is a form of continuing analysis...it sets the agenda for the next wave of data collection" (Miles and Huberman, 1994). Another way to explain it is to say that a "code" is a label (or index word) that we attach to paragraphs or any segments of text so we can find those segments easily. Developing a complex coding scheme can be a useful analytic and retrieval tool for a trained social scientist, but no coding scheme can cover every possible way of looking at textual data without being impossibly large, unwieldy, and complex. While the coding of qualitative data is necessary, it should be minimized when using text management and retrieval software like DtSearch.
If coding of qualitative data is kept to a minimum, the result is a more user-friendly and accessible database. A reduction in the number of codes reduces problems in intra- and inter-coder reliability. As well, there can be a tremendous savings in time and money. The methods of coding described below allow researchers to improve their ability to maintain and simply access a textual database.
Suggested Components of a Minimalist Coding System:
1. Use words instead of numbers or mnemonics, because it will be less confusing. Word codes are more likely to be understood and remembered by more people.2. Do not use words that already appear in the text, they can be easily searched by DtSearch. Thus, lower-level descriptive types of codes may not be necessary.
3. Codes should be used for concepts at high and occasionally medium levels of abstraction (i.e., they should stand for complex concepts or relationships between concepts, such as delay (in care-seeking), decisions (about treatment seeking).
4. While this is only a guiding figure, a good minimalist coding system should have no more than 50 codes. Expansion of the system may be required as the research grows to cover new areas of investigation.
5. Some sort of coding system should be implemented to protect confidentiality of subjects when required.
6. It is imperative that a "code book" be written as part of this coding process, both so that the codes created are recorded, and the things that will be included under each code are as described.
Instructions
Following the guidelines above, go through the field notes and begin devising a list of medium and high-level abstract codes, using Form C2. 1. It may be necessary to generate research questions (Form C3. 1), to help identify relevant codes. Several suggestions can be made on how to enter codes, once they have been devised:
1. Since it should be clear that they are codes, we suggest making them look different from regular text, by capitalizing them, putting in brackets, and by adding a symbol on each end of the code (for instance, (# Status Of Women #)). The extra symbol allows for searches exclusive or inclusive of the code. For ease in reading, it might also be useful to highlight/boldface the codes.2. If a limited number of codes are devised, it may be possible to "hot-key" a number of codes that respond to a set of simple keystrokes. For instance, in WordPerfect macros can be developed using the "alt" key in combination with a letter to print the code. This saves a lot of time in the coding process. Coding systems that require more than 26 codes can consider acquiring microcomputer programs that permit increased numbers of user-definable keys, such as SMARTKEY or PROKEY.
3. I recommend placing the codes immediately following the text they relate to. That way "searches within" are likely to be most effective.
4. Locating appropriate places to insert codes can be a lengthy process involving reading and rereading field notes. DtSearch can be used to hasten this searching process.
Once the preliminary coding process is complete, the notes can be indexed or re-indexed using DtSearch. Preliminary searches of the data should be conducted at this point. If the investigator is still in the field, these preliminary searches can help him or her identify gaps in the data that can be areas for continued research. The coding list should be refined and expanded at this point as well.
FORM C2.1
Generating Research Questions
|
Respondent: |
Date: |
|
Research Question: |
Level Code: |
|
1. | |
|
2. | |
|
3. | |
|
4. | |
|
5. | |
|
6. | |
|
7. | |
|
8. | |
|
9. | |
|
10 | |
|
11. | |
|
12. |
Level Codes:
ST- simple term
SC - simple category
MA - middle-level abstraction
HA - high-level abstraction