
Message/Author 

Mario posted on Monday, January 11, 2016  11:27 pm



We are using MPlus 7.0 and looking for a way to compute the AICc values. We are using ESTIMATOR = MLR and get AIC values. We are wondering if we could use the Burnham & Anderson (2002) equation to convert AIC to AICc : AICc = AIC+(((2*k)*(k+1))/(nk1)) we are asking since this equation was built for normal distributed data whereas in our model some dependent variables have logistic distribution. Also, we are wondering whether for the “k” in the above equation we can use the value “Number of Free Parameters” that MPlus provides. Thanks, Mario 


I'm sorry, but I am not familiar with that B & A equation. You may want to ask the authors. 

Mario posted on Wednesday, January 13, 2016  6:01 am



Thanks for the answer. We are wondering whether there is any statistical reason why MPlus doesn’t provide AICc values. If so, is there another way to compare models with small sample size (e.g., N=141; free parameters~20)? Thanks, Mario 


No statistical reason, except I don't think that BIC is too bad even at n=141. If you really want that measure, just plug in the values and compute it by hand. It does look like their formula uses k for the number of parameters. 

Back to top 

