An Approximate Confidence Interval for the Mean of the Dependent Variable in a Random One-Way Classification with Covariates

Abstract

Consider a dependent variable y linearly related to one or more independent variables, where several parallel lines must be used to model the relationship due to data points occurring in distinct batches or random observation classes. We suggest an approximate confidence interval for the expected value of y at a specified point in the space of the independent variables. The approach stems from a distributional approximation and is supported by a limited simulation.

KEYWORDS:

• Regression
• Conlidence intervals
• Covariates
• Random classes
• Simulation
• One-way classification