An extended logit-normal regression with application to human development index data

Abstract

We propose the odd log-logistic logit-normal regression. Monte Carlo simulations reveal the consistency of the maximum likelihood estimates. The new regression is better than two known regressions to explain the human development index of 478 cities in two Brazilian states.

Keywords:

• Bimodal data
• Maximum likelihood
• Quantile residuals
• Regression models
• Simulation study

Acknowledgment

The authors acknowledge the financial support from CNPq and CAPES, Brazil. CNPq Scholarship – Brazil (166774/2020-0)