Skip to main content
. 2018 Nov 28;38(19):3507–3526. doi: 10.1002/sim.8018

Table 4.

Model details and results of the multilevel modelling

Fractional Goodness of fit
polynomial powers
Model specification Detail N Median SD Deviance Observations Observations
<3rd centile >97th centile
Random intercept Take the mean of the 20 030 2, 2 1 136 690 638 (3.2%) 647 (3.2%)
two‐level model triplicate FHC
measurements
for each visit
Random intercept Take the mean of the 20 030 2, 2 1 132 845 866 (4.3%) 923 (4.6%)
and slope triplicate FHC
two‐level model measurements
for each visit
Random intercept Randomly select 20 030 2, 2 1 135 710 926 (4.6%) 1009 (5.0%)
slope and one of the three FHC
two‐level model measurements
for each visit
Random intercept Consider all 59 973 2, 2 3, 3, 3 346 036 2993 (5.0%) 3231(5.4%)
and slope three data levels
three‐level model