Higher Order Theme Analysis

• Correlation Analysis
• Confirmatory Factor Analysis
• Scale Development
• Regression Analysis

Correlational Analysis

• Analyze for each SCIV theme/topic
• Group items into constructs
• Create item-by-item
• correlation matrix for each construct

Correlational Analysis Violence Against Women (VAW)

Group items which theoretically measure VAW constructs (awareness, knowledge, attitudes, actual practice, social norm)

• Hypothetically, 6 specific items may be grouped together because SC believes theoretically that those items measures knowledge.

Calculate and create correlation matrices for each VAW construct

• Develop a matrix containing bivariate correlations between the 6 items

Confirmatory Factor Analysis

• Analyze for each SCIV theme/topic
• Enter developed item-by-item correlation matrix for each construct into CFA
• Reject items which contain large errors

Confirmatory Factor Analysis (VAW)

Enter each VAW item-by-item construct matrix into CFA

• Input the 6 item correlation matrix for knowledge into CFA

Eliminate high error items for each VAW construct

• Hypothetical, eliminate the 2 items that CFA indicates contain high error in the factor structure

Keep remaining items for each VAW construct

• Relabel VAW knowledge construct with the remaining 4 items.

Scale Development

Analyze for each SCIV theme/topic

Develop new scales/variables for constructs for analysis purposes.

Scale Development (VAW)

For each VAW construct, determine if a new variable/scale must be developed for analysis

Establish criteria for the new variable/scale for analysis purposes

• Hypothetically, reclassify the VAW knowledge construct (4 items) into a new scale score which is a count of correct responses (1 pt for each correct responses), such that the new VAW knowledge scale ranges from 0 (highly unknowledge) to 4 (highly knowledge)

Regression Analysis

Analyze for each SCIV theme/topic

Perform Multivariate Regression or Logistic Regression

Regress demographic variables, confounding variables, and SCIV exposure variables onto scale-developed outcome constructs