Skip to main content

Table 3 LCMM parameter estimates for 1 to 5 classes (N = 273) using a Quadratic Trajectory Function

From: Trajectories of adherence to mood stabilizers in patients with bipolar disorder

N classes Maximum log likelihood AICa BICb Relative entropy Class parameter Class 1 Class 2 Class 3 Class 4 Class 5
1 716.85 − 1419.69 − 1394.43   N
%
APPAc
273
100%
1.000
    
2d 1076.39 − 2130.79 − 2091.09 0.878 N
%
APPA
210
77%
0.971
63
23%
0.968
   
3 1095.99 − 2161.98 − 2107.84 0.889 N
%
APPA
20
7%
0.965
202
74%
0.968
51
19%
0.905
  
4 1095.99 − 2153.98 − 2085.40 0.549 N
%
APPA
20
7%
0.965
198
73%
0.507
0
0%
0.000
55
20%
0.870
 
5 1095.99 − 2145.98 − 2062.97 0.428 N
%
APPA
20
7%
0.965
0
0%
0.000
0
0%
0.000
193
71%
0.344
60
22%
0.824
  1. Trajectory function was weekly adherence = a0 + a1 * week + a2 * week2. No covariates were included
  2. aAkaike Information Criterion
  3. bBayesian Information Criterion
  4. cAverage posterior probability of assignment
  5. dPreferred model