Analyses of childhood stunting have mainly used mean regression yet modeling using quantile regression is more appropriate than using mean regression in that the former provides flexibility to analyze the determinants of stunting corresponding to quantiles of interest whereas the latter allows only analyzing the determinants of mean stunting. Bayesian structured additive quantile regression models were fitted for childhood stunting. Both quantile and mean regression models were fitted and their estimates were compared. Inference was fully Bayesian using integrated nested Laplace approximation approach for quantile regression and Markov chain and Monte Carlo approach for mean regression. The 2010 Malawi demography and health surveys data was used. Using multistage stratified sampling, more than 19000 eligible reproductive women aged between 15 and 49 years were interviewed in a round of surveys and the anthropometric characteristics of their under 5 children were measured. We found that the dominant determinants of childhood stunting in Malawi include child sex, household head sex, type of residence, mother working status, vitamin A supplementation, availability of radio/TV, source of drinking water, vaccination coverage, infectious diseases, mother education, ethnicity, child age, and duration of breastfeeding. We also observed no any significant structured spatial effects on childhood stunting. In this study, we confirmed that quantile regression fits better than mean regression when modeling childhood stunting.