Please Note: This seminar will be held in-person.
|
Student seminar series
Adam
Howes M3 3127 |
Integrated nested Laplace approximations for extended latent Gaussian models with application to the Naomi HIV model
Naomi
(Eaton
et
al,
2021)
is
a
spatial
evidence
synthesis
model
used
to
produce
district-level
HIV
epidemic
indicators
in
sub-Saharan
Africa.
Multiple
outcomes
of
interest,
including
HIV
prevalence,
HIV
incidence
and
treatment
coverage
are
jointly
modelled
using
both
household
survey
data
and
routinely
reported
health
system
data.
The
model
is
provided
as
a
tool
for
countries
to
input
their
data
to
and
generate
estimates
(see
https://naomi.unaids.org/).
In
this
setting,
computationally
intensive
inference
methods
like
MCMC
are
impractical.
To
enable
fast
and
accurate
inference
for
Naomi,
and
other
extended
latent
Gaussian
models,
we
are
developing
a
new
inference
method
which
combines
the
simplified
integrated
nested
Laplace
approximation
approach
of
Wood
(2020)
with
adaptive
Gauss-Hermite
quadrature.
The
new
method
will
be
implemented
as
an
extension
of
the
aghq
package
(Stringer,
2021),
which
will
facilitate
flexible
and
particularly
easy
use
when
provided
a
TMB
C++
template
for
the
log-posterior.
In
this
talk,
I’ll
discuss
this
progress
towards
this
project,
which
I
have
been
working
on
with
Alex
Stringer
here
at
Waterloo
this
term
through
the
International
Visiting
Graduate
Student
program.