Page 23 - Contributed Paper Session (CPS) - Volume 4

CPS2106 Julio M. Singer et al.

This algorithm, adapted from Muggeo et al. (2014), essentially considers
iterative fitting of standard linear mixed models by restricted maximum
likelihood. At convergence, we expect a neglible difference between the third
and fourth terms in the right hand side of (2) and as a consequence, that the
pseudo observations should well approximate the original ones. Given the
linear mixed model nature of the proposed fitting algorithm, we may employ
the diagnostic procedures outlined in Singer et al. (2017) to check whether the
adopted assumptions for the distribution of the random effects or of the
random error are reasonable.
Estimates of the parameters of model (1) obtained via fitting the
approximation (2) along with the corresponding standard errors are
summarized in Table (1).
The results of a Wald test for the homogeneity of the six changepoints ψ1
2
(χ = 58.30,df = 5,p < 0.001) suggests further analyses to identify the possible
effects of treatment, sex and their interaction. A significant interaction
between treatment and sex with respect to the ψ1 changepoints (χ = 13.65,df
2
= 2,p = 0.001) may be analysed via the multiple comparisons summarized in
Table 2 and suggest that the onset of symptoms for the control group (HBSS)
males is delayed by 1.7 [CI(95%) = 1.0, 2.4] weeks with respect to the control
group females and that treatment with Pericytes (both sexes) or MSC (females)
delay the onset of symptoms by 1.3 [CI(95%) = 0.5, 2.2] weeks with respect to
HBSS treated males. The changepoint for MSC treated males lies between
those for HBSS treated males and females but the small sample size does not
lead to a significant difference in either case.
The results for a similar analysis of the acceleration with which the
symptom progresses suggest no difference between sexes and an increase in
2
the acceleration of 18.6 [CI(95%) = 17.8, 19.6] sec/week for the experimental

12 | I S I W S C 2 0 1 9