I have a GLM where I'm predicting a proportion response (# correct items / total # of quiz items). The common recommendation in this case is to assume a binomial distribution and use a logit link. For example: What are the issues with using percentage outcome in linear regression? and Regression for an outcome (ratio or fraction) between 0 and 1

Some others in my department (varying levels of statistical knowledge) disagree, or agree but suggest the use of Poisson or a linear model for simplicity.

In addressing their concerns, and also anticipating reviewer responses to the analysis, I'm wondering if there are any references (e.g. journal articles) that I could cite that suggest/recommend binomial GLM for this type of outcome variable?


Recommended in comments:

McCullagh, P., & Nelder, J. A. (1989). Generalized linear models (Monographs on Statistics and Applied Probability 37). CRC press.

Hardin, J. W. & Hilbe, J. M. (2007). Generalized linear models and extensions. Stata press.

Zhao, L., Chen, Y., & Schaffner, D. W. (2001). Comparison of logistic regression and linear regression in modeling percentage data. Applied and Environmental Microbiology, 67(5), 2129-2135.

  • $\begingroup$ h/t @TheLaconic $\endgroup$ – mkt Aug 21 '18 at 11:58
  • $\begingroup$ h/t @kjetilbhalvorsen $\endgroup$ – mkt Aug 21 '18 at 11:58

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.