Using Markov chain model to evaluate medical students’ trajectory on progress tests and predict USMLE step 1 scores---a retrospective cohort study in one medical school

Table 1 Selection of number of latent states using Latent Markov Model

Number of Latent States	Test score ranges in each latent state	AIC^a	BIC^b
3	[0,45], (45,59), (59,100]	25,851.63	26,014.38
4	[0, 44], (44, 56], (56, 69], (69, 100]	25,875.68	25,991.92
5	[0, 44], (44, 55], (56, 62], (62, 71], (71, 100]	26,197.04	26,243.54
6	[0, 43], (43, 50], (50, 56], (56, 64], (64, 72], (72,100]	25,928.72	26,006.22

^aAIC equals − 2* loglikehood + 2* number of parameters
^bBIC equals − 2* loglikehood + log (number of observations)* number of parameters

ISSN: 1472-6920