In　order　to　gain　a　more　detailed　understanding　of　novice　driver　involved　crashes,　this　paper　compares　the　injury　severity　of　at-fault　and　not-at-fault　novice　driver　involved　crashes　with　the　consideration　of　risky　driving　behaviours.　The　crash　data　of　novice　drivers　in　Shenyang,　China,　from　2012　to　2017　is　used.　The　correlated　random　parameters　logit　model　with　heterogeneity　in　means　is　built　for　the　whole　dataset,　at-fault　dataset,　and　not-at-fault　dataset　to　explore　the　significant　factors　determining　the　injury　severity,　and　obtain　insight　into　hidden　and　common　factors.　The　hidden　factors　indicate　factors　that　are　only　significant　in　one　dataset,　while　the　common　factors　indicate　factors　that　are　significant　in　two　or　more　datasets.　The　results　show　that　nine　common　factors　significantly　affect　injury　severity　of　novice　driver　involved　crashes.　Among　them,　speeding,　legal　but　unsafe　driving　behaviours,　and　improper　operation　are　related　to　risky　driving　behaviour.　Moreover,　road　type,　functional　zone,　time　interval,　and　conflict　and　improper　passing　are　hidden　factors,　among　which　conflict　and　improper　passing　is　related　to　risky　driving　behaviour.　Finally,　some　valuable　insights　for　regulatory　authorities　in　formulating　targeted　policies　are　provided　to　mitigate　the　severity　of　novice　drivers　involved　crashes.