Confidence intervals using contrasts for regression model

Abstract

A graph of confidence intervals can be used to report results from a regression model with explanatory variables asfactors. In this paper we describe a method for computing and displaying confidence intervals using weighted sum contraststo compare population means in unbalanced designs. We extend this method to models with covariates and logistic regression models.

Keywords

• 95% confidence interval
• weighted sum contrasts
• unbalanced designs