Suppose that Y is a dependent variable of interest (such as the survival time or some monotone transformation of the survival time) and x is a vector of covariates. The observed data consist of a random sample from .

You can use regression analysis to explore the relationship between the response Y and its predictor x. Classical linear regression estimates the conditional mean function by using a linear predictor ; a linear quantile regression estimates the th conditional quantile function by using a different linear predictor , where the quantile level ranges between 0 and 1. For example, is the linear predictor for the 0.95th quantile (commonly referred to as the 95th percentile).

The quantile regression coefficient can be estimated by minimizing the following objective function over b:

The loss function is defined as , in contrast to the square loss function for classical linear regression.

When , the coefficient minimizes the sum of absolute residuals, which corresponds to median regression (or regression).

The following set of regression quantiles is referred to as the *quantile process*, and it completely describes the conditional distribution of Y, given the predictor x:

When all the observations are observed, you can use the QUANTREG procedure to estimate the quantile function and draw statistical inference about the regression parameters . For more information, see ChapterĀ 95: The QUANTREG Procedure.

However, when the observations are incomplete, as is the case with censored data in survival analysis, the classical quantile regression method is not appropriate. The QUANTLIFE procedure implements appropriate quantile regression methods to model the relationship between the response Y and the predictor x.