Markov regression model for analyzing big data to predict trajectories of repeated categorical outcomes: an application to $$\hbox {PM}_{2.5}$$ PM 2.5 air pollution data