Nick Irons
@nickirons.bsky.social
Florence Nightingale Fellow
@oxfordstatistics.bsky.social @oxforddemsci.bsky.social
statistics, public health, urbanism, running, art.
njirons.github.io
header @odilonredon.bsky.social
@oxfordstatistics.bsky.social @oxforddemsci.bsky.social
statistics, public health, urbanism, running, art.
njirons.github.io
header @odilonredon.bsky.social
Keefe's Say Nothing is great
December 10, 2024 at 1:52 PM
Keefe's Say Nothing is great
likely at least partially responsible for why the US has become such an antisocial place
December 6, 2024 at 10:08 PM
likely at least partially responsible for why the US has become such an antisocial place
Perhaps just a heavy-tailed Lee-Carter would do the trick as well? instead of using two stages
December 4, 2024 at 8:33 PM
Perhaps just a heavy-tailed Lee-Carter would do the trick as well? instead of using two stages
These posterior projections give an approximation to the full posterior (with uncertainty propagated correctly). This is the idea behind multiple imputation
December 4, 2024 at 7:52 PM
These posterior projections give an approximation to the full posterior (with uncertainty propagated correctly). This is the idea behind multiple imputation
my initial take would be:
(1) fit a bayesian model to the pre-trend (can use a heavy-tailed likelihood to account for outliers). (2) take a random sample of (say 100) posterior trajectories, fitting Lee-Carter to each one and project forward. Each one of these is a posterior projection.
(1) fit a bayesian model to the pre-trend (can use a heavy-tailed likelihood to account for outliers). (2) take a random sample of (say 100) posterior trajectories, fitting Lee-Carter to each one and project forward. Each one of these is a posterior projection.
December 4, 2024 at 7:49 PM
my initial take would be:
(1) fit a bayesian model to the pre-trend (can use a heavy-tailed likelihood to account for outliers). (2) take a random sample of (say 100) posterior trajectories, fitting Lee-Carter to each one and project forward. Each one of these is a posterior projection.
(1) fit a bayesian model to the pre-trend (can use a heavy-tailed likelihood to account for outliers). (2) take a random sample of (say 100) posterior trajectories, fitting Lee-Carter to each one and project forward. Each one of these is a posterior projection.
have you considered going Bayesian
December 4, 2024 at 4:49 PM
have you considered going Bayesian
masking was extremely cost-effective arxiv.org/abs/2411.12016
US COVID-19 school closure was not cost-effective, but other measures were
Non-pharmaceutical interventions (NPIs) in response to the COVID-19 pandemic necessitated a trade-off between the health impacts of viral spread and the social and economic costs of restrictions. We c...
arxiv.org
November 28, 2024 at 10:00 PM
masking was extremely cost-effective arxiv.org/abs/2411.12016