Bias analysis for misclassification in a multicategorical exposure in a logistic regression model

We derive the expressions of asymptotic biases when ignoring the misclassification in a multicategory exposure. For a model with a misclassified exposure variable only, we provide a general conclusion on the direction of the biases under nondifferential misclassification assumption. To better understand the bias formulas, we use a numerical example.