The sudden spin-down in the rotation of magnetar 1E 2259+586 observed by Archibald et al . ( 1 ) was a rare event . However this particular event , referred to as an anti-glitch , was followed by another event which Archibald et al . ( 1 ) suggested could either be a conventional glitch or another anti-glitch . Although there is no accompanied radiation activity or pulse profile change , there is decisive evidence for the existence of the second timing event , judging from the timing data . We apply Bayesian Model Selection to quantitatively determine which of these possibilities better explains the observed data . We show that the observed data strongly supports the presence of two successive anti-glitches with a Bayes Factor , often called the odds ratio , greater than 40 . Furthermore , we show that the second anti-gtlich has an associated frequency change \Delta \nu of -8.2 \times 10 ^ { -8 } Hz . We discuss the implications of these results for possible physical mechanisms behind this anti-glitch .