data.txt

NOFRAME

\newcommand{\jdtitle}{Data, models and science}
\newcommand{\jdsub}{}
\newcommand{\years}{2019--\thisyear}

----------------------------------------------------------------------

ICI3DTHEME

----------------------------------------------------------------------

SEC Introduction

----------------------------------------------------------------------

Goals

	Process of science

	How science informs public health

		Specific examples

	Approaches to epidemiology

----------------------------------------------------------------------

Science is a \emph{process}

	Observe and experiment with reality to \emph{discover} and \emph{challenge}
	ideas about how it works

	A key to science is that everything is open to question

		Science is the belief in the ignorance of experts -- \emph{Feynman}

----------------------------------------------------------------------

The process of science

FIG pearson-15.loop.jpg

----------------------------------------------------------------------

Science without experiments

FIG pearson-16.loop.jpg

----------------------------------------------------------------------

EXTRA NOTES Science with and without experiments

DOUBLEFIG pearson-15.loop.jpg pearson-16.loop.jpg

ADD Did not work (pix not right size)

----------------------------------------------------------------------

SEC Public health

----------------------------------------------------------------------

SS Maternal mortality

----------------------------------------------------------------------

PICSLIDE DIAGRAM maternal.Rout-0.pdf

----------------------------------------------------------------------

Observation and action

	In 1840, medical students stopped visiting Clinic 1

	In 1847, a surgeon died from infection following a scalpel injury

		Igor Semmelweiss made medical students wash their hands

----------------------------------------------------------------------

PICSLIDE DIAGRAM maternal.Rout-0.pdf

----------------------------------------------------------------------

PICSLIDE DIAGRAM maternal.Rout-1.pdf

----------------------------------------------------------------------

DIAGRAM maternal.Rout-2.pdf

----------------------------------------------------------------------

Looking at the data

	ANS Why was Clinic 1 so dangerous in the 40s?

	ANS \ldots and so safe in 1851 and 1852?

	ANS What can we learn from modern statistics?

	ANS And what can't we learn without more data?

----------------------------------------------------------------------

DIAGRAM maternal.Rout-3.pdf

----------------------------------------------------------------------

TSS Cholera

	Is it caused by bad air, or bad water?

	What's bad about it?

----------------------------------------------------------------------

PSLIDE Cholera and air

FIG Farr.Rout-0.pdf

----------------------------------------------------------------------

Cholera and air

FIG Farr.Rout-1.pdf

----------------------------------------------------------------------

PIC FIG snow0.pdf

----------------------------------------------------------------------

PIC FIG snow_pumps.pdf

----------------------------------------------------------------------

TSS Yellow fever and malaria

	Ross determined the cause of malaria primarily by experiments on mosquitoes

	Reed determined the cause of yellow fever primarily by experiments on human
	volunteers

----------------------------------------------------------------------

SEC Approaches to epidemiology

----------------------------------------------------------------------

Data, models and science

	We're never finished, we compare models to data over and over again

	Data is what we use to develop and understand models

	Models are what we use to interpret data

		and they can suggest what data we need to collect

	Complicated or hard-to-test theories may require \emph{dynamical} models

---------------------------------------------------------------------

BCC

\textbf{Classical epidemiology}

PRESENT \vspace{8ex}

	Avoid mechanism

	Control for non-independence of ``units''

PRESENT \vspace{30ex}

NCC

\textbf{Dynamical epidemiology}

PRESENT \vspace{8ex}

	Embrace mechanism

	Explicitly incorporate dependence between units

		X is infected because Y infected them

PRESENT \vspace{30ex}

EC

----------------------------------------------------------------------

SS Classical epidemiology

----------------------------------------------------------------------


Classical example

FIG smoking.Rout-0.pdf

----------------------------------------------------------------------

Classical example

FIG smoking.Rout-1.pdf

----------------------------------------------------------------------

Univariate means

FIG smoke_effects.Rout-0.pdf

----------------------------------------------------------------------

Multivariate means

FIG smoke_effects.Rout-1.pdf

----------------------------------------------------------------------

TSS Dynamical epidemiology

DIAGRAM SIR_model_family/sir.three.pdf

----------------------------------------------------------------------

PSLIDE Polio

WFIG 0.8 webpix/polio_pink.jpg

----------------------------------------------------------------------

PSLIDE Measles

WFIG 0.8 webpix/measles_pink.jpg

----------------------------------------------------------------------

PSLIDE Rubella

WFIG 0.8 webpix/rubella_pink.jpg

----------------------------------------------------------------------

PSLIDE Influenza

WFIG 0.8 webpix/flu_pink.jpg

----------------------------------------------------------------------

NSLIDE Other viruses

Pictures from CDC Pink book \url{https://www.cdc.gov/vaccines/pubs/pinkbook/index.html}

	Rubella

	Measles

	Polio

	Influenza

----------------------------------------------------------------------

Bridging

	Classical epidemiology relies on statistics, avoids mechanism

	Mathematical epidemiology (the traditional approach to dynamical
	epidemiology) explores mechanism, avoids statistics

	Much modern dynamical epidemiology seeks ways to put dynamical mechanisms
	into a statistical framework

		This is hard

----------------------------------------------------------------------

TSS Building knowledge from data

	We must have communication up and down the analysis pipeline

		Data are collected in the field

		Organized and documented

		Protected (for confidentiality, and often for the rights of collectors)

		Summarized

		Modeled

----------------------------------------------------------------------

Example: COVID incidence

	Positive test results but no negative test results

		Can't correct for testing intensity

	Positive and negative test results, but no individual identifiers

		Can't correct for multiple testing of the same people

	Test results, but not reason for testing

		Can't correct for testing focus

----------------------------------------------------------------------

Example: COVID variants

	Mutational screens not linked to invididuals

		We can estimate mutations, but not variants

	Reasons for screening or sequencing not provided

		Can't correct for selection bias

	Personal information not provided

		Can't look for geographical patterns, vaccine effectiveness, \ldots

----------------------------------------------------------------------

Example: West Africa Ebola Outbreak

	Medical vs.~public-health priorities

		Am I responsible if my data request increases the pressure on a
		front-line responder?

	Individual-level vs.~population considerations

		When it is OK to randomize people to receive a placebo vaccine or
		treatment?

----------------------------------------------------------------------

Example: Microbiome studies

	In some fields, the amount of apparently high-quality data is far
	outstripping the ability to understand it

	Lots of reasonably intelligent experiments (or observational designs)

	Huge lists of taxonomic (or metagenomic) communities

----------------------------------------------------------------------

PIC FIG my_images/notindependent.png

CREDIT https://www.nature.com/articles/s41598-021-92400-y

----------------------------------------------------------------------

Data vs.~models

	Models can teach us a lot, but good data with a simplistic model is usually
	better than poor (or poorly contextualized) data with a good model

	Sometimes, the most valuable thing about the model is that it helps us
	figure out what data we need

		ANS Value-of-information models

----------------------------------------------------------------------

TSEC Summary

	Science is an ongoing process

	Models are the way that we bridge between theory and reality

	We can only bridge to reality if we can measure reality

		Collect and curate good data

	Dynamical models have a key role

		When we can't do experiments

		When mechanisms are complex

	We should work to combine dynamics with statistical approaches