Predicting gender differences as latent variables: summed scores, and individual item responses: a methods case study

Area of Science:

• Health outcomes research
• Psychometrics
• Biostatistics

Background:

• Measuring latent variables like physical disability is complex due to reliance on proxy measures.
• Significant methodological challenges arise from using indirect indicators for abstract constructs.

Purpose of the Study:

• To present and compare three distinct methods for predicting latent variables.
• Methods include classical summed scores, individual item responses, and latent variable models.

Main Methods:

• Literature review and data analysis employing a "layers of information" approach.
• Utilized data from the North Carolina Back Pain Project with a modified Roland Questionnaire.

Main Results:

• Initial linear regression showed a gender difference of 1.32 points (95% CI 0.65, 2.00).
• Item analysis yielded inconsistent results; IRT models identified three items with differential item functioning.
• Removing these items reduced the gender difference to 0.78 points (95% CI -0.99, 1.23), with robust findings via re-sampling.

Conclusions:

• The validity of observed differences in latent variables depends on the chosen statistical model.
• Model assumptions can distort true differences, necessitating model comparison for accurate interpretation.