The impact of incomplete data on quantile regression for longitudinal data

Anneleen Verhasselt, Alvaro José Flórez, Ingrid Van Keilegom, Geert Molenberghs

Keywords: Inverse probability weighting, Longitudinal data, Missing data, Pseudo-likelihood, Quantile regression

Abstract: We investigate the performance of quantile methods for longitudinal data with missingness. In a simulation study, we compare the performance of the quantile regression using different alternatives for handling missing data and taking the correlation into account. As expected, the non-likelihood-based methods provide biased estimates under the missing at random assumption. On the other hand, an inverse probability weighting approach corrects for biasedness.