Skip to main content

Table 2 Marginal AIC values of models where time axis is numerical. In the case of models with optimized knot location(s), we show the AIC’s of models with the lowest AIC amongst models of the same complexity with different knot locations. All AIC’s calculated in this case can be seen in the Supplementary Tables and are calculated using ML estimation. The cells with an NA show unbuilt models. This is because knot optimization in 2-segment first order polynomial models of these adjuvants did not result in much gain, unlike the last two adjuvants (see first and third rows), therefore we skipped knot optimization in the 3-segment models for those. Models in the last two rows were again built for all adjuvants due to their expected good performance

From: Modeling adaptive response profiles in a vaccine clinical trial

Fixed Effect Structure

Random Effect Structure

Model specification

Adjuvant

   

AS01B

AS01E

AS03A

AS04

Alum

2-segment 1st order polynomial

2 slopes

knot at Day 44

1179.27

1228.55

1189.45

1254.54

827.64

3-segment 1st order polynomial

3 slopes

knots at Days 30 & 44

767.99

796.35

723.94

865.89

675.69

2-segment 1st order polynomial

2 slopes

knot optimized

1175.51

1223.95

1179.99

1216.55

784.10

3-segment 1st order polynomial

3 slopes

knots optimized

NA

NA

NA

865.89

669.64

3rd order polynomial

all degrees except intercept

x-axis transformed

1385.02

1389.77

1344.64

1327.85

822.13

2-segment 2nd order polynomial

all degrees except intercept

x-axis transformed, knot optimized

1189.33

1227.23

1181.72

1210.15

765.78