4058.0: Tuesday, November 14, 2000 - 9:00 AM

Abstract #3748

Intimate partner violence agreement III: Measures of agreement for ordinal violence data

Jeffrey A. Welge, PhD, Department of Psychiatry and Center for Biostatistical Services, University of Cincinnati Medical Center, P.O. Box 670559, 231 Bethesda Avenue, Cincinnati, OH 45267-0559, (513)558-0856, welgeja@email.uc.edu and John Schafer, PhD, Department of Psychology, University of Cincinnati, ML 0376, Cincinnati, OH 45221-0376.

When both partners of a dyad respond to questions about intimate partner violence (IPV) having >2 ordered response categories, a number of candidate measures of agreement might be considered. These can be roughly arranged on a continuum from one-dimensional summary measures (e.g., Cohen’s kappa and Pearson’s r) to full-information measures (e.g., distribution of Partner B’s response, given Partner A’s response). Summary measures provide the most concise description of agreement, but are sensitive to the marginal distribution of the responses. Decomposing agreement into multiple components (e.g., agreement about violence and agreement about non-violence) and using graphical displays similar in spirit to the Q-Q distribution plot are promising intermediate strategies. Graphical presentation of the distribution of one partner’s responses at each level of the other partner’s response and computation of conditional misclassification rates is the most informative approach, but naturally imposes the greatest informational load on data analysts and consumers. We illustrate the use of several agreement measures, using a frequency-of-violence item from the National Survey of Families and Households.

Question: In the last year, how many fights resulted in your partner hitting/ shoving/throwing things at you?

B

A

None

One

Two

Three

Four or more

Marginal Total

None

20

9

5

1

0

35

One

3

26

12

8

2

51

Two

2

10

9

4

4

29

Three

1

5

6

6

5

23

Four or more

1

4

2

5

6

18

Marginal Total

27

54

34

24

17

156

Learning Objectives: Participants will learn to identify the trade-offs involved in choosing measures of agreement and will be introduced to a number of standard and novel agreement indices

Keywords: Methodology, Domestic Violence

Presenting author's disclosure statement:
Organization/institution whose products or services will be discussed: None
I do not have any significant financial interest/arrangement or affiliation with any organization/institution whose products or services are being discussed in this session.

The 128th Annual Meeting of APHA