This study develops a sparse network model and changepoint detection framework to identify drivers and changes in users' posting probability on social networks. With the model, we examine the impact of user behaviour on the Reddit forum Wallstreetbets on the stock price of GameStop (GME). Results show that changepoints in vital users' behaviour significantly predicted GME's integrated volatility with a time lag and non-vital users behaviour immediately, even when controlling for network activity and established metrics measuring influential user impact. This suggests that the change in users behaviour was more important for the GME market frenzy than the joint activity of the Wallstreetbets users. Likewise return dynamics and jump volatility can be explained with change in users behaviour. We conclude that the activity of a small group of vital nodes, impacting the behaviour of other nodes, on Wallstreetbets drove the activity of the GME traders.