methods-structured-interviewing

Anthropac — Freelists

Adapted from the v. 3.22 user manual
Chapter 2-1

FREE LISTS

The FREELIST procedure reads free list data and counts the number of times each item is mentioned. FREELIST uses the Soundex algorithm (see <http://www.myatt.demon.co.uk/sxalg.htm>) to identify items that are spelled differently but sound alike.

Methodology

The free list is used to identify the principal items in a cognitive domain. Essentially, you ask several informants a question like "What are all the pets you can think of?". The resulting lists of items are typed into a file and tabulated. The most frequently mentioned items are then explored further, perhaps through the use of pilesorts, triads tests, or ratings tasks.

Data Format

For each informant, the data should begin with a pound sign (#) and an optional identifier (such as the informant's name or id number), followed by the items he or she mentioned. Each item should be typed on a line by itself. For example, the following file contains hypothetical free lists of the birth control domain from 3 respondents, identified as "Father Mulcahey", "Tammy Faye Bakker", and "Jim Bakker".

Contents of file PETS.FL:

# Bob
cats
dogs
fish
snake
bird
# Tammy
dog
cat
gerbil
hamster
bird
ferret
# Jim
dog
horse
cat
frog

Cleaning Up the Data

The first thing that Anthropac does is walk us through the data item by item identifying those that appear similar to others already mentioned. Upon finding similar items, the program asks whether we want to recode them so that they are the same. Processing the above data, for example, would yield this query:

FREE LIST
-------------------------------------------
Trying to match DOG:

2 DOGS

(DOG) Enter #, -#, A, D, R or ? for help:

If we type "?"<Enter> we learn that the commands are as follows

If DOG matches an item on the list above
then enter the number of the item that DOG matches
or minus (-) the number to make DOG the new
spelling.
else enter A to add DOG to the list of items, or
D to ignore DOG altogether.

We ask ourselves whether we want the standard coding to be singular or plural and, supposing we decide on singular (in practice we'd go through the process more than once or look over our data to see which is most common) and so enter "-2" and press enter.

Outputs

SORTED FREQUENCIES. A list of items mentioned, sorted in descending order by frequency, is output to the screen (unless redirected via the OUTPUT procedure).

ITEM FREQUENCY RESP PCT
-------------------------------------------
1 DOG 3 100
2 CAT 3 100
3 BIRD 2 67
4 FISH 1 33
5 GERBIL 1 33
6 HAMSTER 1 33
7 FROG 1 33
8 HORSE 1 33
9 FERRET 1 33
10 SNAKE 1 33
-------------------------------------------
Total mentions; mentions per respondent: 15 5.000

Cleaned up freelist data written to file CLEANED.DAT
Respondent-by-item matrix written to file FLMATRIX.DAT
A copy of frequency distribution has been saved in file FREQ.OUT

CLEANED UP FILE. A copy of the data file in which spelling differences have been resolved is written to disk.

# BOB
CATS
DOGS
FISH
SNAKE
BIRD
# TAMMY
DOG
CAT
GERBIL
HAMSTER
BIRD
FERRET
# JIM
DOG
HORSE
CAT
FROG

RESPONDENT-ITEM MATRIX. A binary respondent-by-item matrix is also written to disk. The Xij cell of the matrix is 1 if respondent i mentions item j and 0 otherwise. For example, in the case of the file BCONTROL.FL, the following output file is created:

Contents of file FLMATRIX.DAT:

DL NR = 3 NC = 10
ROW LABELS:
"BOB"
"TAMMY"
"JIM"
COLUMN LABELS:
"CAT"
"DOG"
"FISH"
"SNAKE"
"BIRD"
"GERBIL"
"HAMSTER"
"FERRET"
"HORSE"
"FROG"
DATA:
1 1 1 1 1 0 0 0 0 0
1 1 0 0 1 1 1 1 0 0
1 1 0 0 0 0 0 0 1 1