cts401 analyzing and interpreting data from the revised conflict tactics scales and the...
Post on 17-Dec-2015
219 Views
Preview:
TRANSCRIPT
CTS40 1
ANALYZING AND INTERPRETING DATA FROM THE REVISED CONFLICT TACTICS SCALES
AND THE INTERNATIONAL DATING VIOLENCE STUDY
Murray A. StrausFamily Research Laboratory, University of New Hampshire
Durham, NH 03824 603-862-2594 murray.straus@unh.edu
Many papers on the CTS are available on my website: http://pubpages.unh.edu/~mas2
Workshop presented at the International Conference on Family Violence Research,Portsmouth, NH 15 July 2003.
CTS40 2
This workshop assumes at least some familiarity with the CTS. It focuses on the following topics (1) The conceptual and operational difference between the four versions of each scale, for example, the Physical Assault Scale provides separate scale scores for the Frequency, Chronicity, Prevalence, and Ever Prevalence of assault. (2) The types of issues that are best investigated with each of the four versions and why the Prevalence score is the most frequently used type of score. (3) Ways to take into account the severity of assault, injury, psychological aggression, and sexual coercion, including the advantages and limitations of weighted scores. (4) An explanation and discussion of the tables in a paper comparing the results from use of the CTS2 in Juarez Mexico and New Hampshire, USA. (6) An explanation and discussion of the tables in a paper testing the hypothesis that corporal punishment is related to violence against a dating partner. (5) A discussion period in which participants in the workshop bring up questions concerning the use and interpretation of their data. Approximately half the workshop time will be reserved for the discussion period. The questions will be answered in ways that will be useful for all participants in the workshop, not just the person who posed the question.
CTS40 3
TOPICS COVERED
1. The multiplicity of CTS scales and scores and how to manage them
2. When to use the Frequency, Chronicity, Prevalence, and Ever Prevalence scores
3. How to take into account severity of assault, injury etc
4. Explanation of tables from a paper comparing Juarez Mexico and New Hampshire, USA
5. Explanation of tables from a paper on the hypothesis that corporal punishment is related to violence against a dating partner
CTS40 4
THE FIVE CTS2 SCALES* Physical Assault* Injury* Sexual Coercion* Psychological Aggression* Negotiation
Each as subscales for “severe” and “minor” (less severe) levels
CTS40 5
FOUR WAYS OF SCORING EACH CTS SCALE
Prevalence (short for “annual prevalence”)One or more occurrences of any of the acts in the scaleScore range 0 - 1
FrequencyThe number or occurrences of the acts in a CTS scaleScore range 0 - (25* the number of items in the scale). For a six
item scale this would be 0-150Meaning of a score of 12 or 25
ChronicityThe number of occurrences of the acts in a CTS scale by a partner who has done at least one of the acts Meaning of a score of 6, 12, or25
Ever PrevalenceOne or more occurrences of any of the acts in the scale since the
relationship beganUsually best to avoid because of memory problems
CTS40 6
DON’T LET THE EXTENSIVE DATA FROM THE CTS2TEMPT YOU TO WRITE AN ARTICLE THAT IS
UNDIGESTIABLE BECAUSE IT HAS TOO MUCH IN IT
• 5 SCALES, EACH WITH 2 OR MORE SUBSCALES, AND BOTH VICTIMIZATION AND PERPETRATION DATA
= 5 * 2 * 2 = 20 variables• If you include the prevalence rate AND the chronicity for each scale:
= 20 * 2 = 40 variables • It is almost always important to present results for males and females separately,
which further multiples the amount of information in an article
TWO STRATEGIES• Restrict the paper to either perpetration or victimization unless you have an important reason to compare them• Write separate papers for each of the scales unless you want to compare scores on
two scales, or you have some other important reason to include two or more
CTS40 7
ISSUES BEST INVESTIGATED WITH EACH OF THE FOUR SCORES
PREVALENCEA. What percent of perpetrated or were victimized? (can also be a rate per 1,000,
10,000 etc)* Among groups in your study such as young and old, treatment &
control* In your sample compared with other populations
FREQUENCYHow often did the behavior occur in the past year? (or whatever referent period was
used)* Not appropriate with general population samples because the mean
and even the median is misleading. * Appropriate with batterer treatment participants or victim samples
CHRONICITYIn a general population, among the sub-set who manifest the behavior, how often
did it occur?EVER PREVALENCE
If the behavior did not occur in the past year, was there victimization or perpetration before that?
CTS40 8
WHY PREVALENCE IS THE MOST FREQUENTLY USED SCORE
•Everyone understands a percentage
•Makes for easy comparison of subgroups
•Does no exacerbate the skewed distribution and outlier problem
•Can be used with logistic regression to obtain odds ratios and predicted probabilities
Even when other scores are presented, it is wise to also present the prevalence rates. It lets readers know where your population is.
CTS40 9
HOW TO TAKE INTO ACCOUNT SEVERITY OF ASSAULT, INJURY, SEXUAL COERCION, AND PSYCHOLOGICAL AGGRESSION
•USE THE SEVERE SUBSCALE(For Sexual Coercion, the Threat and Force subscales)
* USE THE FREQUENCY OR CHRONICITY SCOREMeasures severity by how often the behavior occurs
* COMPUTE A “SEVERITY WEIGHTED” FREQUENCY OR CHRONICITY SCOREBUT: In a general population sample, exacerbate skewness and
outlier problems Useful for issues such as measuring change in severity of
perpetration or victimization
• USE SEVERITY-LEVEL CATEGORIES0 = None, 1 = Minor Only, 2 = SevereCan be the dependent variable in multinomial logistic regression. This gives
separate coefficients for Minor Only and for SevereMy current recommendation
CTS40 10
MODES OF ANALYSIS
FREQUENCY DISTRIBUTIONCROSS TABANOVA AND ANCOVAREGRESSION
OLSROBUSTLOGISTIC – BINARY AND MULTINOMIAL
FACTOR ANALYSISGENDER SPECIFIC ANALYSEMULTILEVEL MODELING
CTS40 11
TABLE FROM A PAPER ON VIOLENCE AGAINST DATING PARTNERS BY UNIVERSITY STUDENTS IN JUAREZ MEXICO
Ignacio Luis RamirezTexas Technological Universiy
STATISTICAL METHODS USED TO CREATE THE TABLES
INTERPRETING THE TABLES
CTS40 12
EXAMPLE OFTOO MUCH IN ONE ARTICLE
Table X. Prevalence and chronicity statistics by gender of participant * _________________________________________________________ Gender of Participant** Scale Men Women __________________________________________________________
A. Tactic "Expressed," "Enacted," "Perpetrated," “Inflicted”
Negotiation Prevalence (%) 100 98 Chronicity (Mean) 61.6 69.7 (SD) 38.5 38.8
Psychological Aggression Prevalence (%) 74 83 Chronicity (Mean) 15.1 16.0 (SD) 17.4 18.8
Physical Assault Prevalence (%) 47 35 Chronicity (Mean) 12.9 9.4 (SD) 21.6 15.4
Sexual Coercion Prevalence (%) 37 18 Chronicity (Mean) 19.9 12.6 (SD) 31.4 15.8
Injury revalence (%) 15 9 Chronicity (Mean) 25.1 3.6 (SD) 37.8 6.8
B. Tactic "Experienced," "Received,”
"Victimization,” “Sustained” Negotiation
Prevalence (%) 100 99 Chronicity (Mean) 57.4 67.1 (SD) 35.5 39.7
Psychological Aggression revalence (%) 76 78 Chronicity (Mean) 17.2 15.1 (SD) 21.1 18.5
Physical Assault Prevalence (%) 49 31 Chronicity (Mean) 15.9 9.3 (SD) 25.8 18.0
Sexual Coercion Prevalence (%) 38 30 Chronicity (Mean) 18.5 11.8 (SD) 30.2 14.4
Injury male partners, as reported by men I Prevalence (%) 16 14 Chronicity (Mean) 24.7 6.2 (SD) 41.7 11.6
_________________________________________________________
CTS40 13
Table X. Prevalence and chronicity statistics by gender of participant * _________________________________________________________ Gender of Participant** Scale Men Women __________________________________________________________
A. Tactic "Expressed," "Enacted," "Perpetrated," “Inflicted”
Negotiation Prevalence (%) 100 98 Chronicity (Mean) 61.6 69.7 (SD) 38.5 38.8
Psychological Aggression Prevalence (%) 74 83 Chronicity (Mean) 15.1 16.0 (SD) 17.4 18.8
Physical Assault Prevalence (%) 47 35 Chronicity (Mean) 12.9 9.4 (SD) 21.6 15.4
Sexual Coercion Prevalence (%) 37 18 Chronicity (Mean) 19.9 12.6 (SD) 31.4 15.8
Injury revalence (%) 15 9 Chronicity (Mean) 25.1 3.6
(SD) 37.8 6.8
THE TOP HALF OF THE TABLE
CTS40 14
Table X. Prevalence and chronicity statistics by gender of participant * _______________________________________________________ Gender of Participant** Scale Men Women _______________________________________________________
A. Tactic "Expressed," "Enacted," "Perpetrated," “Inflicted”
Physical Assault Prevalence (%) 47 35 Chi-sq = Chronicity (Mean) 12.9 9.4 F = (SD) 21.6 15.4
Injury Prevalence (%) 15 9 Chi-sq = Chronicity (Mean) 25.1 3.6 F =
(SD) 37.8 6.8
Two scales is probably as much as should be in one paper because it is usually important to give separate results for:
Severe level (not in the above table)Both Prevalence and Chronicity (as above)
INTERPRETATION OF DATA IN THIS TABLE
CTS40 15
TABLES FROM A PAPER ON THE HYPOTHESIS THAT CORPORAL PUNISHMENT IS RELATED TO VIOLENCE
AGAINST A DATING PARTNER
Emily Douglas and Murray A. StrausUniversity of New Hampshire
WILL DISCUSS
• STATISTICAL METHODS USED TO CREATE THE TABLES
• INTERPRETING THE TABLES
CTS40 16
Table 2: Rates of Attitudes Related to Violence And Actual Violence, By Gender Of Student
Male% Female% Total %
Hit a lot before the age of 12 58.17 48.69 51.91 Hit a lot as a teenager 34.36 25.75 28.63 Appropriate for husband to hit wife 49.39 42.63 44.93 Appropriate for wife to hit husband 75.76 76.48 76.55 Total assault against a partner 25.73 30.14 28.99 Severe assault against a partner 9.65 11.35 10.96 Total injury perpetrated against a partner 8.66 7.41 7.78 Severe injury perpetrated against a partner 4.02 2.2 2.83
CTS Rates
These are CTS prevalence rates
* Gender differences can be tested using cross tabs and chi-square. * Using ANOVA or ANCOVA lets you control for other variables and test for interactions, but with a dichotomous dependent variable, it violates the assumptions.
CTS40 17
SPANKDED OR HIT A LOT BEFORE AGE 12
Strongly AgreeAgreeDisagreeStrongly Disagree
Perc
ent W
ho In
jure
d A
Partn
er
14%
12%
10%
8%
6%
4%
2%
0%
Figure 4. SEVERE INJURY OF DATING PARTNERBY CORPORAL PUNISHMENT AS A CHILD
AND GENDER OF STUDENT ANALYSIS OF COVARIANCEDEPENDENT VARIABLE• CTS Injury Prevalence, coded 0-100INDEPDENDENT VARIABLES• Corporal punishment before age 12• Gender of studentINTERACTION• Corporal punishment with GenderCOVARIATES (controls)• Age of Student• Length of Relationship• Socioeconomic Status of Family• Social Desirability ScaleOPTIONS• Plot • Effect Size (gives ETA squared)
This violates assumptions of ANCOVA.So does Frequency score because it is so skewed. Prevalence is the best alternative.
Males
Females
F = 4.76, p <.01
CTS40 18
LOGISTIC REGRESSION IS OFTEN THE BEST APPROACH
* Dependent variable can be dichotomous (binomial) or multinomial* Dichotomous: CTS Prevalence scores (% perpetrating or % victimized)* Multinomial examples: Severity Level (None, Minor Only, Severe)
Couple Type (None, Male Only, Female Only, Both)
•Allows for: Multiple independent variables and control variables Gives the“net effect” of each independent variable
* Provides results in two forms that are more widely understood than regression coefficients:
Odds ratio. Example: “The odds ratio of 1.087 indicates that each increase of one point in the seven category corporal punishment index increases the odds of a man physically assaulting his partner by 8.7%” (Straus & Yodanis, 1996)
Predicted Probability of the dependent variable occurring (see example slide)
* USE GRAPHS as in Straus, Murray A. and Carrie L. Yodanis. 1996. "Corporal punishment in adolescence and physical assaults on spouses later in life: What accounts for the link?" Journal of Marriage and the Family 58:825-841.
CTS40 19
CTS40 20
CTS40 21
top related