Quantile Regression Analysis; Simulation Study With Violation of Normality Assumption

Quantile regression is an extension method of simple linear regression whose work is to separate or divide data into certain quantiles. This method minimizes the asymmetric absolute residual and estimates the conditional quantile function. Parameter estimation in the quantile regression method does not require the parametric assumption of normality. The data in this study are generated from different distributions. The distribution of the independent variables in this study comes from the t distribution, normal and exponential distribution. Meanwhile, the error distribution comes from the chi square distribution. This research produces various models of the selected quantiles. The estimated parameter values at each quantile are almost close to the initial values set. This research found the best model at quantile 0.5 by looking at the smallest MSE value of all quantiles of 1.2662. The best model obtained is .